Update config.yaml

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

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）
 

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: "v1.6.42"
+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
 # ============================================
@@ -93,13 +96,75 @@ fofa:
 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）
   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
@@ -109,7 +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
@@ -126,7 +192,7 @@ 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）
+    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
@@ -144,6 +210,7 @@ multi_agent:
     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
+    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

internal/agent/agent.go

@@ -779,13 +779,26 @@ 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）。

internal/agent/default_single_system_prompt.go

@@ -113,5 +113,7 @@ func DefaultSingleAgentSystemPrompt() string {
 
 - 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

@@ -21,10 +21,13 @@ 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"
@@ -66,6 +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()
 
@@ -99,12 +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)
@@ -129,6 +145,10 @@ func New(cfg *config.Config, log *logger.Logger, configPath string) (*App, error
 		externalMCPMgr.StartAllEnabled()
 	}
 
+	execReconciler := monitor.NewExecutionReconciler(db, mcpServer, externalMCPMgr, log.Logger)
+	execReconciler.ReconcileOnStartup()
+	monitor.StartStaleRunningReconcileLoop(execReconciler, log.Logger)
+
 	// 创建Agent
 	maxIterations := cfg.Agent.MaxIterations
 	if maxIterations <= 0 {
@@ -298,7 +318,9 @@ func New(cfg *config.Config, log *logger.Logger, configPath string) (*App, error
 	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)
-	db.SetEinoConversationDirs(plantaskBase, checkpointBase)
+	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
@@ -325,7 +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)
@@ -343,6 +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)
@@ -368,6 +394,7 @@ 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, conversationHandler, agentHandler)
@@ -791,11 +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)

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
@@ -95,9 +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.
@@ -106,6 +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 {
@@ -249,7 +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"`
@@ -263,12 +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"`
 }
@@ -489,6 +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"`
@@ -604,15 +631,109 @@ type DatabaseConfig struct {
 type AgentConfig struct {
 	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 {
@@ -623,6 +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.
@@ -782,33 +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
@@ -852,6 +970,10 @@ func Load(path string) (*Config, error) {
 		}
 	}
 
+	if err := ValidateWecomConfig(cfg.Robots.Wecom); err != nil {
+		return nil, err
+	}
+
 	return &cfg, nil
 }
 
@@ -1076,6 +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
@@ -1252,8 +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
@@ -1274,6 +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/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)
@@ -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)

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)
@@ -395,7 +423,7 @@ func conversationOrderClause(sortBy, tableAlias string) string {
 }
 
 // ListConversations 列出所有对话
-func (db *DB) ListConversations(limit, offset int, search, sortBy string) ([]*Conversation, error) {
+func (db *DB) ListConversations(limit, offset int, search, sortBy, projectID string) ([]*Conversation, error) {
 	var rows *sql.Rows
 	var err error
 
@@ -403,20 +431,30 @@ func (db *DB) ListConversations(limit, offset int, search, sortBy string) ([]*Co
 		// 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 ?)
+			 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 "+orderClause+" LIMIT ? OFFSET ?",
-			limit, offset,
+			"SELECT id, title, COALESCE(pinned, 0), created_at, updated_at, project_id FROM conversations"+where+" "+orderClause+" LIMIT ? OFFSET ?",
+			args...,
 		)
 	}
 
@@ -472,23 +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, sortBy string) ([]*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+`
+			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)
@@ -533,6 +578,19 @@ func (db *DB) ListUngroupedConversations(limit, offset int, sortBy string) ([]*C
 	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
@@ -585,12 +643,14 @@ func (db *DB) DeleteConversation(id string) error {
 		// 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)
 	}
-	db.removeConversationScopedDirs(id)
+	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
@@ -628,13 +688,50 @@ func (db *DB) removeConversationScopedDir(base, conversationID, label string) {
 	}
 }
 
-func (db *DB) removeConversationScopedDirs(conversationID string) {
-	// summarization transcript, reduction files, etc.
+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。
@@ -811,6 +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) {
@@ -918,6 +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()
@@ -979,6 +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

@@ -19,7 +19,9 @@ func TestDeleteConversationRemovesEinoScopedDirs(t *testing.T) {
 
 	plantaskBase := filepath.Join(tmp, "skills", ".eino", "plantask")
 	checkpointBase := filepath.Join(tmp, "eino-checkpoints")
-	db.SetEinoConversationDirs(plantaskBase, checkpointBase)
+	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 {
@@ -34,6 +36,8 @@ func TestDeleteConversationRemovesEinoScopedDirs(t *testing.T) {
 		{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 {
@@ -48,10 +52,57 @@ func TestDeleteConversationRemovesEinoScopedDirs(t *testing.T) {
 		t.Fatalf("DeleteConversation: %v", err)
 	}
 
-	for _, base := range []string{db.conversationArtifactsDir, plantaskBase, checkpointBase} {
+	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/database.go

@@ -51,6 +51,8 @@ type DB struct {
 	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{}
@@ -159,12 +161,16 @@ func NewDB(dbPath string, logger *zap.Logger) (*DB, error) {
 
 // SetEinoConversationDirs configures best-effort filesystem cleanup on DeleteConversation.
 // plantaskBase is skills_root/plantask_rel (no conversation id); checkpointBase is checkpoint_dir root.
-func (db *DB) SetEinoConversationDirs(plantaskBase, checkpointBase string) {
+// 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
@@ -405,6 +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,
@@ -1134,6 +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
 }
 

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"
 
@@ -227,6 +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 := `
@@ -288,6 +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 = ?`
@@ -410,6 +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
@@ -530,13 +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 {
@@ -544,35 +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
 }
 

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/execute 桥在工具调用结束时 Fire，
-// 用于清除 pending tool_call（tool_result 由 ADK schema.Tool 事件推送，含流式工具与 reduction 后正文）。
+// 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"`
 }
@@ -631,40 +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
-	var emptyResponseAttempts int
-	for {
-		resultMA, errMA = multiagent.RunEinoSingleChatModelAgent(
-			taskCtx, h.config, &h.config.MultiAgent, h.agent, h.db, h.logger,
-			conversationID, h.conversationProjectID(conversationID), curMsg, curHist, roleTools, progressCallback, nil, h.projectBlackboardBlock(conversationID),
-		)
-		handledEmpty, exhaustedEmpty := h.handleEinoEmptyResponseContinue(
-			taskCtx, conversationID, resultMA, errMA, &emptyResponseAttempts,
-			&curHist, &curMsg, segmentUserMessage, progressCallback, nil,
-		)
-		if exhaustedEmpty {
-			errMA = nil
-			break
-		}
-		if handledEmpty {
-			continue
-		}
-		if errMA == nil {
-			transientRunAttempts = 0
-			emptyResponseAttempts = 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)
 	}
@@ -680,41 +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
-	var emptyResponseAttempts int
-	for {
-		resultMA, errMA = multiagent.RunDeepAgent(
-			taskCtx, h.config, &h.config.MultiAgent, h.agent, h.db, h.logger,
-			conversationID, h.conversationProjectID(conversationID), curMsg, curHist, roleTools, progressCallback,
-			h.agentsMarkdownDir, orchestration, nil, h.projectBlackboardBlock(conversationID),
-		)
-		handledEmpty, exhaustedEmpty := h.handleEinoEmptyResponseContinue(
-			taskCtx, conversationID, resultMA, errMA, &emptyResponseAttempts,
-			&curHist, &curMsg, segmentUserMessage, progressCallback, nil,
-		)
-		if exhaustedEmpty {
-			errMA = nil
-			break
-		}
-		if handledEmpty {
-			continue
-		}
-		if errMA == nil {
-			transientRunAttempts = 0
-			emptyResponseAttempts = 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)
 	}
@@ -879,11 +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
@@ -896,6 +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
@@ -915,6 +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()
 	}
@@ -951,6 +940,7 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 			}
 			flushedThinking[sid] = true
 		}
+		syncHitlCognition()
 	}
 
 	return func(eventType, message string, data interface{}) {
@@ -1011,6 +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 {
@@ -1218,6 +1227,7 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 					respPlan.meta[k] = v
 				}
 			}
+			syncHitlCognition()
 			return
 		}
 		if eventType == "response" {
@@ -1287,6 +1297,7 @@ func (h *AgentHandler) createProgressCallback(runCtx context.Context, cancelRun
 					}
 				}
 			}
+			syncHitlCognition()
 			return
 		}
 
@@ -1338,10 +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"`
 	}
@@ -1356,27 +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。
@@ -1408,6 +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))
@@ -1474,17 +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,
 	})
 }
 
@@ -1498,6 +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。
@@ -1557,7 +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
@@ -1747,15 +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
 	}
@@ -1830,9 +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 {
@@ -1926,7 +2026,7 @@ func (h *AgentHandler) RunSingleBatchTask(c *gin.Context) {
 
 	// 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.forceUnmarkBatchQueueRunning(queueID)
+		h.batchTaskManager.ForceUnmarkQueueExecutor(queueID)
 	}
 
 	autoStarted := true
@@ -1985,26 +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) forceUnmarkBatchQueueRunning(queueID string) {
-	h.unmarkBatchQueueRunning(queueID)
-}
-
 func (h *AgentHandler) nextBatchQueueRunAt(cronExpr string, from time.Time) (*time.Time, error) {
 	expr := strings.TrimSpace(cronExpr)
 	if expr == "" {
@@ -2020,43 +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())
@@ -2108,324 +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)
-			if h.batchTaskManager.TakeSingleRunTaskIfMatch(queueID, task.ID) {
-				h.batchTaskManager.UpdateQueueStatus(queueID, "paused")
-				break
-			}
-			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.db, h.logger, conversationID, h.conversationProjectID(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.db, h.logger, conversationID, h.conversationProjectID(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)
-
-		if h.batchTaskManager.TakeSingleRunTaskIfMatch(queueID, task.ID) {
-			h.batchTaskManager.UpdateQueueStatus(queueID, "paused")
-			h.logger.Info("单条执行完成，队列已暂停", zap.String("queueId", queueID), zap.String("taskId", task.ID))
-			break
-		}
-
-		// 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/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"`
@@ -80,8 +97,9 @@ 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
+	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
 }
 
@@ -93,11 +111,56 @@ func NewBatchTaskManager(logger *zap.Logger) *BatchTaskManager {
 	return &BatchTaskManager{
 		logger:         logger,
 		queues:         make(map[string]*BatchTaskQueue),
-		taskCancels:    make(map[string]context.CancelFunc),
+		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()
@@ -105,10 +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
@@ -136,6 +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(),
@@ -177,6 +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))
@@ -272,6 +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
 	}
@@ -511,6 +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
 		}
@@ -651,8 +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)
 	}
@@ -680,9 +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))
 		}
 	}
@@ -868,7 +958,6 @@ func (m *BatchTaskManager) AddTaskToQueue(queueID, message string) (*BatchTask,
 
 // 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 cancelFunc context.CancelFunc
 	var siblingRunningIDs []string
 
 	m.mu.Lock()
@@ -898,11 +987,9 @@ func (m *BatchTaskManager) PrepareSingleTaskRun(queueID, taskID string) error {
 	}
 
 	// 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 {
-		if c, ok := m.taskCancels[queueID]; ok {
-			cancelFunc = c
-			delete(m.taskCancels, queueID)
-		}
+		cancelFuncs = m.drainTaskCancelsLocked(queueID)
 		for _, t := range queue.Tasks {
 			if t != nil && t.ID != taskID && t.Status == BatchTaskStatusRunning {
 				siblingRunningIDs = append(siblingRunningIDs, t.ID)
@@ -914,8 +1001,10 @@ func (m *BatchTaskManager) PrepareSingleTaskRun(queueID, taskID string) error {
 	resumeQueue := queue.Status == BatchQueueStatusCompleted || queue.Status == BatchQueueStatusCancelled
 	m.mu.Unlock()
 
-	if cancelFunc != nil {
-		cancelFunc()
+	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 {
@@ -1089,7 +1178,90 @@ func queueAllowsSingleTaskRunLocked(queue *BatchTaskQueue, task *BatchTask) bool
 	}
 }
 
-// GetNextTask 获取下一个待执行的任务
+// 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()
@@ -1130,20 +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]
@@ -1168,17 +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
@@ -1187,7 +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]
@@ -1228,34 +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
@@ -1269,7 +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/config.go

@@ -798,6 +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),
@@ -1329,6 +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")
 
@@ -1751,6 +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()
@@ -1771,6 +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) {

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)
@@ -103,7 +115,7 @@ 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"))
 
@@ -111,14 +123,14 @@ func (h *ConversationHandler) ListConversations(c *gin.Context) {
 	var total int
 	var err error
 	if excludeGrouped {
-		conversations, err = h.db.ListUngroupedConversations(limit, offset, sortBy)
+		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, sortBy)
+		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 {
@@ -165,6 +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 == "" {
@@ -172,6 +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))
@@ -180,14 +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{}{
@@ -200,8 +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
@@ -245,6 +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

@@ -2,31 +2,11 @@ package handler
 
 import (
 	"context"
-	"errors"
-	"fmt"
-	"strings"
-	"time"
 
 	"cyberstrike-ai/internal/agent"
 	"cyberstrike-ai/internal/multiagent"
-
-	"go.uber.org/zap"
 )
 
-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,
@@ -45,136 +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
-}
-
-// handleEinoEmptyResponseContinue 在 SSE 任务循环内处理「正常结束但无助手正文」；返回 exhausted=true 时由外层按成功结束（保留占位文案）。
-// 与临时错误重试一致：仅恢复轨迹并保留本请求原始 user 文案，不向模型注入续跑说明。
-func (h *AgentHandler) handleEinoEmptyResponseContinue(
-	baseCtx context.Context,
-	conversationID string,
-	result *multiagent.RunResult,
-	runErr error,
-	emptyResponseAttempts *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, exhausted bool) {
-	if !errors.Is(runErr, multiagent.ErrEmptyResponseContinue) {
-		return false, false
-	}
-	maxAttempts := h.einoRunRetryMaxAttempts()
-	*emptyResponseAttempts++
-	if *emptyResponseAttempts > maxAttempts {
-		if h.logger != nil {
-			h.logger.Warn("eino empty response auto resume exhausted",
-				zap.String("conversationId", conversationID),
-				zap.Int("maxAttempts", maxAttempts))
-		}
-		if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
-			h.persistEinoAgentTraceForResume(conversationID, result)
-		}
-		return false, true
-	}
-	attemptNo := *emptyResponseAttempts
-	if h.logger != nil {
-		h.logger.Info("eino empty response, auto resume from trace",
-			zap.String("conversationId", conversationID),
-			zap.Int("attempt", attemptNo),
-			zap.Int("maxAttempts", maxAttempts))
-	}
-	if progressCallback != nil {
-		progressCallback("eino_empty_response_continue", fmt.Sprintf("未捕获到助手正文，正在基于轨迹自动续跑（%d/%d）…", attemptNo, maxAttempts), map[string]interface{}{
-			"conversationId": conversationID,
-			"source":         "eino",
-			"attempt":        attemptNo,
-			"maxAttempts":    maxAttempts,
-			"resumeKind":     "trace_segment",
-		})
-	}
-	h.applyEinoTransientRetrySegment(conversationID, result, curHistory, curFinalMessage, segmentUserMessage)
-	if sendProgress != nil {
-		sendProgress("已恢复上下文，正在继续推理…", map[string]interface{}{
-			"conversationId": conversationID,
-			"source":         "empty_response_continue",
-		})
-	}
-	return true, false
-}

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
 
@@ -177,10 +176,9 @@ func (h *AgentHandler) EinoSingleAgentLoopStream(c *gin.Context) {
 	taskOwned = true
 
 	var cumulativeMCPExecutionIDs []string
-	var transientRunAttempts int
-	var emptyResponseAttempts 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
@@ -215,6 +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)
 		})
@@ -233,61 +232,25 @@ func (h *AgentHandler) EinoSingleAgentLoopStream(c *gin.Context) {
 			roleTools,
 			progressCallback,
 			chatReasoningToClientIntent(req.Reasoning),
-			h.projectBlackboardBlock(conversationID),
+			h.agentSessionContextBlock(conversationID),
 		)
 
 		if result != nil && len(result.MCPExecutionIDs) > 0 {
 			cumulativeMCPExecutionIDs = mergeMCPExecutionIDLists(cumulativeMCPExecutionIDs, result.MCPExecutionIDs)
 		}
 
-		handledEmpty, exhaustedEmpty := h.handleEinoEmptyResponseContinue(
-			baseCtx, conversationID, result, runErr, &emptyResponseAttempts,
-			&curHistory, &curFinalMessage, segmentUserMessage, progressCallback,
-			func(msg string, extra map[string]interface{}) { sendEvent("progress", msg, extra) },
-		)
-		if exhaustedEmpty {
-			runErr = nil
-			transientRunAttempts = 0
-			timeoutCancel()
-			break
-		}
-		if handledEmpty {
-			mainIterationOffset += segmentMainIterationMax
-			transientRunAttempts = 0
-			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 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
-			emptyResponseAttempts = 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) {
@@ -312,8 +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)
@@ -448,8 +409,6 @@ func (h *AgentHandler) EinoSingleAgentLoop(c *gin.Context) {
 	curMsg := prep.FinalMessage
 	var result *multiagent.RunResult
 	var runErr error
-	var transientRunAttempts int
-	var emptyResponseAttempts int
 	for {
 		result, runErr = multiagent.RunEinoSingleChatModelAgent(
 			taskCtx,
@@ -465,30 +424,11 @@ func (h *AgentHandler) EinoSingleAgentLoop(c *gin.Context) {
 			prep.RoleTools,
 			progressCallback,
 			chatReasoningToClientIntent(req.Reasoning),
-			h.projectBlackboardBlock(prep.ConversationID),
+			h.agentSessionContextBlock(prep.ConversationID),
 		)
-		handledEmpty, exhaustedEmpty := h.handleEinoEmptyResponseContinue(
-			baseCtx, prep.ConversationID, result, runErr, &emptyResponseAttempts,
-			&curHist, &curMsg, prep.FinalMessage, progressCallback, nil,
-		)
-		if exhaustedEmpty {
-			runErr = nil
-			break
-		}
-		if handledEmpty {
-			continue
-		}
 		if runErr == nil {
 			break
 		}
-		if handled, fatalErr := h.handleEinoTransientRetryContinue(
-			baseCtx, prep.ConversationID, result, runErr, &transientRunAttempts,
-			&curHist, &curMsg, prep.FinalMessage, progressCallback, nil,
-		); handled {
-			continue
-		} else if fatalErr != nil {
-			runErr = fatalErr
-		}
 		if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
 			h.persistEinoAgentTraceForResume(prep.ConversationID, result)
 		}

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
@@ -80,8 +120,9 @@ func (h *MonitorHandler) Monitor(c *gin.Context) {
 	// 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()
@@ -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

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)
@@ -187,10 +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
-	var emptyResponseAttempts 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
@@ -225,6 +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)
 		})
@@ -245,61 +244,25 @@ 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 {
 			cumulativeMCPExecutionIDs = mergeMCPExecutionIDLists(cumulativeMCPExecutionIDs, result.MCPExecutionIDs)
 		}
 
-		handledEmpty, exhaustedEmpty := h.handleEinoEmptyResponseContinue(
-			baseCtx, conversationID, result, runErr, &emptyResponseAttempts,
-			&curHistory, &curFinalMessage, segmentUserMessage, progressCallback,
-			func(msg string, extra map[string]interface{}) { sendEvent("progress", msg, extra) },
-		)
-		if exhaustedEmpty {
-			runErr = nil
-			transientRunAttempts = 0
-			timeoutCancel()
-			break
-		}
-		if handledEmpty {
-			mainIterationOffset += segmentMainIterationMax
-			transientRunAttempts = 0
-			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 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
-			emptyResponseAttempts = 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) {
@@ -324,8 +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)
@@ -460,8 +421,6 @@ func (h *AgentHandler) MultiAgentLoop(c *gin.Context) {
 	curMsg := prep.FinalMessage
 	var result *multiagent.RunResult
 	var runErr error
-	var transientRunAttempts int
-	var emptyResponseAttempts int
 	for {
 		result, runErr = multiagent.RunDeepAgent(
 			taskCtx,
@@ -479,30 +438,11 @@ func (h *AgentHandler) MultiAgentLoop(c *gin.Context) {
 			h.agentsMarkdownDir,
 			strings.TrimSpace(req.Orchestration),
 			chatReasoningToClientIntent(req.Reasoning),
-			h.projectBlackboardBlock(prep.ConversationID),
+			h.agentSessionContextBlock(prep.ConversationID),
 		)
-		handledEmpty, exhaustedEmpty := h.handleEinoEmptyResponseContinue(
-			baseCtx, prep.ConversationID, result, runErr, &emptyResponseAttempts,
-			&curHist, &curMsg, prep.FinalMessage, progressCallback, nil,
-		)
-		if exhaustedEmpty {
-			runErr = nil
-			break
-		}
-		if handledEmpty {
-			continue
-		}
 		if runErr == nil {
 			break
 		}
-		if handled, fatalErr := h.handleEinoTransientRetryContinue(
-			baseCtx, prep.ConversationID, result, runErr, &transientRunAttempts,
-			&curHist, &curMsg, prep.FinalMessage, progressCallback, nil,
-		); handled {
-			continue
-		} else if fatalErr != nil {
-			runErr = fatalErr
-		}
 		if shouldPersistEinoAgentTraceAfterRunError(baseCtx) {
 			h.persistEinoAgentTraceForResume(prep.ConversationID, result)
 		}

internal/handler/openapi.go

@@ -740,14 +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",
@@ -756,20 +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{}{
@@ -1232,6 +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{}{

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 {
@@ -31,6 +70,29 @@ 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 {

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/external_manager.go

@@ -814,6 +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()

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

@@ -921,9 +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 ""
 	}
@@ -931,21 +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"
@@ -953,15 +1004,31 @@ 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）。
@@ -1103,6 +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
+}

internal/multiagent/adk_init.go

@@ -0,0 +1,16 @@
+package multiagent
+
+import (
+	"fmt"
+
+	"github.com/cloudwego/eino/adk"
+)
+
+// InitADK configures global Eino ADK settings. Call once at process startup before
+// any ADK middleware or agents are created.
+func InitADK() error {
+	if err := adk.SetLanguage(adk.LanguageChinese); err != nil {
+		return fmt.Errorf("adk set language: %w", err)
+	}
+	return nil
+}

internal/multiagent/continuation_user_dedup_middleware.go

@@ -0,0 +1,104 @@
+package multiagent
+
+import (
+	"context"
+	"strings"
+
+	"github.com/cloudwego/eino/adk"
+	"github.com/cloudwego/eino/schema"
+	"go.uber.org/zap"
+)
+
+// continuationSessionMarker matches Cursor / IDE session-resume user injections.
+const continuationSessionMarker = "This session is being continued from a previous conversation"
+
+// continuationUserDedupMiddleware keeps only the latest session-resume user message when
+// multiple continuation injections were stacked (e.g. after repeated out-of-context resumes).
+type continuationUserDedupMiddleware struct {
+	adk.BaseChatModelAgentMiddleware
+	logger *zap.Logger
+	phase  string
+}
+
+func newContinuationUserDedupMiddleware(logger *zap.Logger, phase string) adk.ChatModelAgentMiddleware {
+	return &continuationUserDedupMiddleware{logger: logger, phase: phase}
+}
+
+func (m *continuationUserDedupMiddleware) BeforeModelRewriteState(
+	ctx context.Context,
+	state *adk.ChatModelAgentState,
+	mc *adk.ModelContext,
+) (context.Context, *adk.ChatModelAgentState, error) {
+	_ = mc
+	if m == nil || state == nil || len(state.Messages) == 0 {
+		return ctx, state, nil
+	}
+	deduped, dropped := dedupContinuationUserMessages(state.Messages)
+	if dropped == 0 {
+		return ctx, state, nil
+	}
+	if m.logger != nil {
+		m.logger.Info("eino continuation user messages deduplicated",
+			zap.String("phase", m.phase),
+			zap.Int("dropped", dropped),
+			zap.Int("messages_before", len(state.Messages)),
+			zap.Int("messages_after", len(deduped)),
+		)
+	}
+	out := *state
+	out.Messages = deduped
+	return ctx, &out, nil
+}
+
+func adkUserMessageText(msg adk.Message) string {
+	if msg == nil {
+		return ""
+	}
+	var b strings.Builder
+	if s := strings.TrimSpace(msg.Content); s != "" {
+		b.WriteString(s)
+	}
+	for _, part := range msg.UserInputMultiContent {
+		if part.Type == schema.ChatMessagePartTypeText {
+			if s := strings.TrimSpace(part.Text); s != "" {
+				if b.Len() > 0 {
+					b.WriteByte('\n')
+				}
+				b.WriteString(s)
+			}
+		}
+	}
+	return b.String()
+}
+
+func isContinuationUserMessage(msg adk.Message) bool {
+	if msg == nil || msg.Role != schema.User {
+		return false
+	}
+	return strings.Contains(adkUserMessageText(msg), continuationSessionMarker)
+}
+
+func dedupContinuationUserMessages(msgs []adk.Message) ([]adk.Message, int) {
+	lastIdx := -1
+	contCount := 0
+	for i, msg := range msgs {
+		if !isContinuationUserMessage(msg) {
+			continue
+		}
+		contCount++
+		lastIdx = i
+	}
+	if contCount <= 1 {
+		return msgs, 0
+	}
+	out := make([]adk.Message, 0, len(msgs)-(contCount-1))
+	dropped := 0
+	for i, msg := range msgs {
+		if isContinuationUserMessage(msg) && i != lastIdx {
+			dropped++
+			continue
+		}
+		out = append(out, msg)
+	}
+	return out, dropped
+}

internal/multiagent/continuation_user_dedup_middleware_test.go

@@ -0,0 +1,65 @@
+package multiagent
+
+import (
+	"context"
+	"strings"
+	"testing"
+
+	"github.com/cloudwego/eino/adk"
+	"github.com/cloudwego/eino/schema"
+)
+
+func continuationUser(text string) adk.Message {
+	return &schema.Message{
+		Role: schema.User,
+		UserInputMultiContent: []schema.MessageInputPart{
+			{Type: schema.ChatMessagePartTypeText, Text: continuationSessionMarker + "\n" + text},
+			{Type: schema.ChatMessagePartTypeText, Text: "Please continue the conversation from where we left it off."},
+		},
+	}
+}
+
+func TestDedupContinuationUserMessages_KeepsLatest(t *testing.T) {
+	msgs := []adk.Message{
+		continuationUser("summary old"),
+		schema.UserMessage("real task"),
+		continuationUser("summary new"),
+	}
+	out, dropped := dedupContinuationUserMessages(msgs)
+	if dropped != 1 {
+		t.Fatalf("dropped=%d want 1", dropped)
+	}
+	if len(out) != 2 {
+		t.Fatalf("len=%d want 2", len(out))
+	}
+	if out[0].Role != schema.User || adkUserMessageText(out[0]) != "real task" {
+		t.Fatalf("first should remain real task, got %q", adkUserMessageText(out[0]))
+	}
+	if !strings.Contains(adkUserMessageText(out[1]), "summary new") {
+		t.Fatalf("latest continuation not kept: %q", adkUserMessageText(out[1]))
+	}
+}
+
+func TestDedupContinuationUserMessages_NoOpSingle(t *testing.T) {
+	msgs := []adk.Message{continuationUser("only"), schema.UserMessage("task")}
+	out, dropped := dedupContinuationUserMessages(msgs)
+	if dropped != 0 || len(out) != 2 {
+		t.Fatalf("unexpected change dropped=%d len=%d", dropped, len(out))
+	}
+}
+
+func TestContinuationUserDedupMiddleware(t *testing.T) {
+	mw := newContinuationUserDedupMiddleware(nil, "test")
+	state := &adk.ChatModelAgentState{Messages: []adk.Message{
+		continuationUser("old"),
+		continuationUser("new"),
+		schema.UserMessage("task"),
+	}}
+	_, out, err := mw.(*continuationUserDedupMiddleware).BeforeModelRewriteState(context.Background(), state, nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(out.Messages) != 2 {
+		t.Fatalf("want 2 messages after dedup, got %d", len(out.Messages))
+	}
+}

internal/multiagent/eino_adk_run_loop.go

@@ -18,6 +18,7 @@ import (
 	"cyberstrike-ai/internal/einomcp"
 	"cyberstrike-ai/internal/einoobserve"
 	"cyberstrike-ai/internal/openai"
+	"cyberstrike-ai/internal/security"
 
 	"github.com/cloudwego/eino/adk"
 	"github.com/cloudwego/eino/schema"
@@ -90,7 +91,7 @@ type einoADKRunLoopArgs struct {
 	FilesystemMonitorRecord einomcp.ExecutionRecorder
 	MCPExecutionBinder      *MCPExecutionBinder
 
-	// ToolInvokeNotify 与 einomcp.ToolsFromDefinitions 共享：run loop 在迭代前 Set，MCP 桥 Fire 以补全 tool_result。
+	// ToolInvokeNotify 与 einomcp.ToolsFromDefinitions 共享：run loop 在迭代前 Set，execute/MCP 桥 Fire 时立即推送 tool_result（ADK 晚到经 toolResultSent 去重）。
 	ToolInvokeNotify *einomcp.ToolInvokeNotifyHolder
 
 	DA adk.Agent
@@ -196,6 +197,16 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 		pendingByID[tc.ToolCallID] = tc
 		pendingQueueByAgent[tc.EinoAgent] = append(pendingQueueByAgent[tc.EinoAgent], tc.ToolCallID)
 	}
+	markPendingWithMonitor := func(tc toolCallPendingInfo) {
+		markPending(tc)
+		beginEinoADKFilesystemToolMonitor(
+			args.FilesystemMonitorAgent,
+			args.FilesystemMonitorRecord,
+			args.MCPExecutionBinder,
+			tc.ToolCallID,
+			tc.ToolName,
+		)
+	}
 	popNextPendingForAgent := func(agentName string) (toolCallPendingInfo, bool) {
 		pendingMu.Lock()
 		defer pendingMu.Unlock()
@@ -288,6 +299,8 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 
 	var toolResultSent sync.Map // toolCallID -> struct{}；ADK Tool 事件去重（权威正文来自 reduction 处理后的 agent 上下文）
 	tryEmitToolResultProgress := func(toolName, content, toolCallID string, isErr bool, agentName string) {
+		// 仅由 ADK schema.Tool 事件调用；MCP/execute 桥在 reduction 前的 ToolInvokeNotify 不得推送 tool_result，
+		// 否则全量输出会先占位并触发 toolResultSent 去重，导致 UI/监控展示与 agent 实际收到的截断正文不一致。
 		if progress == nil {
 			return
 		}
@@ -305,6 +318,7 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 			"isError":        isErr,
 			"result":         content,
 			"resultPreview":  preview,
+			"agentFacing":    true, // 与 reduction 后送入 ChatModel 的正文一致，供前端展示
 			"conversationId": conversationID,
 			"einoAgent":      agentName,
 			"einoRole":       einoRoleTag(agentName),
@@ -331,7 +345,7 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 			toolCallID = tid
 		}
 		recordPendingExecuteStdoutDup(toolName, content, isErr)
-		recordEinoADKFilesystemToolMonitor(args.FilesystemMonitorAgent, args.FilesystemMonitorRecord, toolName, toolCallID, runAccumulatedMsgs, content, isErr)
+		recordEinoADKFilesystemToolMonitor(args.FilesystemMonitorAgent, args.FilesystemMonitorRecord, args.MCPExecutionBinder, toolName, toolCallID, runAccumulatedMsgs, content, isErr)
 		if args.FilesystemMonitorAgent != nil && args.MCPExecutionBinder != nil {
 			if execID := args.MCPExecutionBinder.ExecutionID(toolCallID); execID != "" {
 				args.FilesystemMonitorAgent.UpdateMCPExecutionDisplayResult(execID, content)
@@ -339,12 +353,6 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 		}
 		progress("tool_result", fmt.Sprintf("工具结果 (%s)", toolName), data)
 	}
-	if args.ToolInvokeNotify != nil {
-		args.ToolInvokeNotify.Set(func(toolCallID, toolName, einoAgent string, success bool, content string, invokeErr error) {
-			removePendingByID(strings.TrimSpace(toolCallID))
-			// tool_result 仅由下方 ADK schema.Tool 事件推送，正文与送入模型的上下文一致（含 reduction 截断）。
-		})
-	}
 
 	if args.EinoCallbacks != nil {
 		ctx = einoobserve.AttachAgentRunCallbacks(ctx, args.EinoCallbacks, einoobserve.Params{
@@ -383,6 +391,12 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 		}
 	}
 	runner := adk.NewRunner(ctx, runnerCfg)
+	startRunnerIter := func(runMsgs []adk.Message) *adk.AsyncIterator[*adk.AgentEvent] {
+		if checkPointID != "" {
+			return runner.Run(ctx, runMsgs, adk.WithCheckPointID(checkPointID))
+		}
+		return runner.Run(ctx, runMsgs)
+	}
 	var iter *adk.AsyncIterator[*adk.AgentEvent]
 	if cpStore != nil && checkPointID != "" {
 		if _, existed, getErr := cpStore.Get(ctx, checkPointID); getErr != nil {
@@ -422,12 +436,9 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 		}
 	}
 	if iter == nil {
-		if checkPointID != "" {
-			iter = runner.Run(ctx, msgs, adk.WithCheckPointID(checkPointID))
-		} else {
-			iter = runner.Run(ctx, msgs)
-		}
+		iter = startRunnerIter(msgs)
 	}
+	transientRetrier := newEinoTransientRunRetrier(einoTransientRunRetryPolicyFromArgs(args))
 	handleRunErr := func(runErr error) error {
 		if runErr == nil {
 			return nil
@@ -480,26 +491,67 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 		return runErr
 	}
 
-	// maybeRetryTransientRun：不在此层 runner.Run/Resume；由 handler 落库 + loadHistoryFromAgentTrace 分段续跑（同中断并继续）。
-	maybeRetryTransientRun := func(runErr error) (retry bool, fatal error) {
-		if runErr == nil || !isEinoTransientRunError(runErr) {
+	maybeRetryTransientRun := func(runErr error) (restarted bool, fatal error) {
+		if runErr == nil {
+			return false, nil
+		}
+		if !isEinoTransientRunError(runErr) {
 			return false, handleRunErr(runErr)
 		}
+		restarted, restartMsgs, ctxSource, backoff, retErr := transientRetrier.tryRetry(
+			ctx, runErr, args, baseMsgs, runAccumulatedMsgs, baseAccumulatedCount,
+		)
+		if retErr != nil {
+			flushAllPendingAsFailed(runErr)
+			if logger != nil {
+				logger.Warn("eino transient retry exhausted",
+					zap.Error(retErr),
+					zap.String("orchestration", orchMode),
+					zap.Int("maxAttempts", transientRetrier.maxAttempts()))
+			}
+			return false, retErr
+		}
+		if !restarted {
+			return false, nil
+		}
+		attemptNo := transientRetrier.attempt()
+		maxAttempts := transientRetrier.maxAttempts()
 		if logger != nil {
-			logger.Warn("eino transient error, ending run segment for handler resume",
+			logger.Warn("eino transient error, retrying after backoff",
 				zap.Error(runErr),
-				zap.String("orchestration", orchMode))
+				zap.String("orchestration", orchMode),
+				zap.Int("attempt", attemptNo),
+				zap.Int("maxAttempts", maxAttempts),
+				zap.Duration("backoff", backoff))
 		}
 		if progress != nil {
-			progress("eino_run_retry", "遇到临时错误（限流或网络波动），将保存上下文并重试…", map[string]interface{}{
+			progress("eino_run_retry", fmt.Sprintf("遇到临时错误（限流或网络波动），%d 秒后第 %d/%d 次重试…", int(backoff.Seconds()), attemptNo, maxAttempts), map[string]interface{}{
 				"conversationId": conversationID,
 				"source":         "eino",
 				"orchestration":  orchMode,
 				"error":          runErr.Error(),
-				"resumeKind":     "trace_segment",
+				"attempt":        attemptNo,
+				"maxAttempts":    maxAttempts,
+				"backoffSec":     int(backoff.Seconds()),
+			})
+			progress("eino_run_retry", "已恢复上下文，正在重试…", map[string]interface{}{
+				"conversationId": conversationID,
+				"source":         "eino",
+				"orchestration":  orchMode,
+				"attempt":        attemptNo,
+				"contextSource":  string(ctxSource),
 			})
 		}
-		return false, ErrTransientRetryContinue
+		msgs = restartMsgs
+		iter = startRunnerIter(msgs)
+		return true, nil
+	}
+
+	// 仅在退避重试后真正收到数据/完成一步时清零，避免重启后首个无错 ADK 事件误把计数打回 0。
+	confirmTransientRetryRecovery := func() {
+		if transientRetrier.attempt() > 0 {
+			transientRetrier.reset()
+		}
 	}
 
 	takePartial := func(runErr error) (*RunResult, error) {
@@ -514,10 +566,10 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 	}
 
 	for {
-		// 检测 context 取消（用户关闭浏览器、请求超时等），flush pending 工具状态避免 UI 卡在 "执行中"。
-		select {
-		case <-ctx.Done():
-			flushAllPendingAsFailed(ctx.Err())
+		// iter.Next 可能长时间阻塞（工具执行、模型推理）；须与 ctx 联动，否则取消/超时无法及时 flush pending。
+		ev, ok, iterCtxErr := nextAgentEventWithContext(ctx, iter)
+		if iterCtxErr != nil {
+			flushAllPendingAsFailed(iterCtxErr)
 			if progress != nil {
 				if isInterruptContinue(ctx) {
 					progress("progress", "已暂停当前输出，正在合并用户补充并继续…", map[string]interface{}{
@@ -526,17 +578,14 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 						"kind":           "interrupt_continue",
 					})
 				} else {
-					progress("error", "Request cancelled / 请求已取消", map[string]interface{}{
+					progress("error", iterCtxErr.Error(), map[string]interface{}{
 						"conversationId": conversationID,
 						"source":         "eino",
 					})
 				}
 			}
-			return takePartial(ctx.Err())
-		default:
+			return takePartial(iterCtxErr)
 		}
-
-		ev, ok := iter.Next()
 		if !ok {
 			// iter 结束并不总是“正常完成”：
 			// 当取消/超时发生在 iter.Next() 阻塞期间时，可能直接返回 !ok。
@@ -583,9 +632,13 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 			continue
 		}
 		if ev.Err != nil {
-			if _, retErr := maybeRetryTransientRun(ev.Err); retErr != nil {
+			restarted, retErr := maybeRetryTransientRun(ev.Err)
+			if retErr != nil {
 				return takePartial(retErr)
 			}
+			if restarted {
+				continue
+			}
 		}
 		if ev.AgentName != "" && progress != nil {
 			iterEinoAgent := orchestratorName
@@ -648,34 +701,9 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 
 		if mv.IsStreaming && mv.MessageStream != nil && mv.Role == schema.Tool {
 			toolName := strings.TrimSpace(mv.ToolName)
-			var toolBuf strings.Builder
-			streamToolCallID := ""
-			var toolStreamRecvErr error
-			for {
-				chunk, rerr := mv.MessageStream.Recv()
-				if errors.Is(rerr, io.EOF) {
-					break
-				}
-				if rerr != nil {
-					toolStreamRecvErr = rerr
-					break
-				}
-				if chunk == nil {
-					continue
-				}
-				if chunk.Content != "" {
-					toolBuf.WriteString(chunk.Content)
-				}
-				if tid := strings.TrimSpace(chunk.ToolCallID); tid != "" {
-					streamToolCallID = tid
-				}
-			}
-			content := toolBuf.String()
-			isErr := false
-			if strings.HasPrefix(content, einomcp.ToolErrorPrefix) {
-				isErr = true
-				content = strings.TrimPrefix(content, einomcp.ToolErrorPrefix)
-			}
+			content, streamToolCallID, toolStreamRecvErr := recvSchemaMessageStream(ctx, mv.MessageStream)
+			isErr := einoToolResultIsError(toolName, content)
+			content = einoToolResultBody(content)
 			if streamToolCallID != "" {
 				opts := []schema.ToolMessageOption{schema.WithToolName(toolName)}
 				runAccumulatedMsgs = append(runAccumulatedMsgs, schema.ToolMessage(content, streamToolCallID, opts...))
@@ -687,6 +715,9 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 					zap.String("agent", ev.AgentName),
 					zap.String("tool", toolName))
 			}
+			if toolStreamRecvErr == nil {
+				confirmTransientRetryRecovery()
+			}
 			continue
 		}
 
@@ -934,7 +965,7 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 			if merged := mergeStreamingToolCallFragments(toolStreamFragments); len(merged) > 0 {
 				lastToolChunk = mergeMessageToolCalls(&schema.Message{ToolCalls: merged})
 			}
-			tryEmitToolCallsOnce(lastToolChunk, ev.AgentName, orchestratorName, conversationID, orchMode, progress, toolEmitSeen, subAgentToolStep, mainAgentToolStep, markPending)
+			tryEmitToolCallsOnce(lastToolChunk, ev.AgentName, orchestratorName, conversationID, orchMode, progress, toolEmitSeen, subAgentToolStep, mainAgentToolStep, markPendingWithMonitor)
 			// 流式路径此前只把 tool_calls 推给进度 UI，未写入 runAccumulatedMsgs；落库后 loadHistory→RepairOrphan 会删掉全部 tool 结果，表现为「续跑/下轮失忆」。
 			if lastToolChunk != nil && len(lastToolChunk.ToolCalls) > 0 {
 				runAccumulatedMsgs = append(runAccumulatedMsgs, schema.AssistantMessage("", lastToolChunk.ToolCalls))
@@ -951,9 +982,15 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 						"einoRole":       einoRoleTag(ev.AgentName),
 					})
 				}
-				if _, retErr := maybeRetryTransientRun(streamRecvErr); retErr != nil {
+				restarted, retErr := maybeRetryTransientRun(streamRecvErr)
+				if retErr != nil {
 					return takePartial(retErr)
 				}
+				if restarted {
+					continue
+				}
+			} else {
+				confirmTransientRetryRecovery()
 			}
 			continue
 		}
@@ -963,7 +1000,7 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 			continue
 		}
 		runAccumulatedMsgs = append(runAccumulatedMsgs, msg)
-		tryEmitToolCallsOnce(mergeMessageToolCalls(msg), ev.AgentName, orchestratorName, conversationID, orchMode, progress, toolEmitSeen, subAgentToolStep, mainAgentToolStep, markPending)
+		tryEmitToolCallsOnce(mergeMessageToolCalls(msg), ev.AgentName, orchestratorName, conversationID, orchMode, progress, toolEmitSeen, subAgentToolStep, mainAgentToolStep, markPendingWithMonitor)
 
 		if mv.Role == schema.Assistant {
 			if progress != nil && strings.TrimSpace(msg.ReasoningContent) != "" {
@@ -1038,15 +1075,13 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 			}
 
 			content := msg.Content
-			isErr := false
-			if strings.HasPrefix(content, einomcp.ToolErrorPrefix) {
-				isErr = true
-				content = strings.TrimPrefix(content, einomcp.ToolErrorPrefix)
-			}
+			isErr := einoToolResultIsError(toolName, content)
+			content = einoToolResultBody(content)
 
 			toolCallID := strings.TrimSpace(msg.ToolCallID)
 			tryEmitToolResultProgress(toolName, content, toolCallID, isErr, ev.AgentName)
 		}
+		confirmTransientRetryRecovery()
 	}
 
 	mcpIDsMu.Lock()
@@ -1057,32 +1092,9 @@ func runEinoADKAgentLoop(ctx context.Context, args *einoADKRunLoopArgs, baseMsgs
 		orchMode, runAccumulatedMsgs, persistTraceSource(args, runAccumulatedMsgs),
 		lastAssistant, lastPlanExecuteExecutor, emptyHint, ids, false,
 	)
-	if shouldEinoEmptyResponseContinue(out, emptyHint, len(runAccumulatedMsgs), baseAccumulatedCount) {
-		if logger != nil {
-			logger.Info("eino empty response, ending run segment for handler resume",
-				zap.String("conversationId", conversationID),
-				zap.String("orchestration", orchMode),
-				zap.Int("traceMessages", len(runAccumulatedMsgs)))
-		}
-		if progress != nil {
-			progress("eino_empty_response_continue", "会话已结束但未产生助手正文，正在基于轨迹自动续跑…", map[string]interface{}{
-				"conversationId": conversationID,
-				"source":         "eino",
-				"resumeKind":     "trace_segment",
-			})
-		}
-		return out, ErrEmptyResponseContinue
-	}
 	return out, nil
 }
 
-func shouldEinoEmptyResponseContinue(out *RunResult, emptyHint string, accumulatedLen, baseCount int) bool {
-	if out == nil || accumulatedLen <= baseCount {
-		return false
-	}
-	return strings.TrimSpace(out.Response) == strings.TrimSpace(emptyHint)
-}
-
 func persistTraceSource(args *einoADKRunLoopArgs, fallback []adk.Message) []adk.Message {
 	if args != nil && args.ModelFacingTrace != nil {
 		if snap := args.ModelFacingTrace.Snapshot(); len(snap) > 0 {
@@ -1097,17 +1109,119 @@ func einoPartialRunLastOutputHint() string {
 		"[Run ended abnormally; continue from the trace above without repeating completed steps.]"
 }
 
-// friendlyEinoExecuteInvokeTail 将 Eino execute 等非 MCP 路径的结尾错误转成简短提示；其它情况保留原 error 文本。
+// friendlyEinoExecuteInvokeTail 将 Eino execute 超时/中断/流异常转为简短提示。
+// 命令非零退出（ExecuteExitError）已有 exec 对齐的正文，不再追加「执行未正常结束」。
 func friendlyEinoExecuteInvokeTail(invokeErr error) string {
 	if invokeErr == nil {
 		return ""
 	}
+	var exitErr *ExecuteExitError
+	if errors.As(invokeErr, &exitErr) {
+		return ""
+	}
 	if errors.Is(invokeErr, context.DeadlineExceeded) {
 		return einoExecuteTimeoutUserHint()
 	}
+	if errors.Is(invokeErr, context.Canceled) {
+		return ""
+	}
+	if strings.Contains(invokeErr.Error(), "shell inactivity timeout") {
+		return ""
+	}
 	return "[执行未正常结束] " + invokeErr.Error()
 }
 
+// einoToolResultIsError 统一判断 Eino 工具结果是否应标记为错误（与 MCP exec 的 IsError 对齐）。
+func einoToolResultIsError(toolName, content string) bool {
+	if strings.HasPrefix(content, einomcp.ToolErrorPrefix) {
+		return true
+	}
+	if strings.TrimSpace(toolName) == "execute" && security.IsCommandFailureResult(content) {
+		return true
+	}
+	return false
+}
+
+// einoToolResultBody 去掉工具错误前缀，返回展示/持久化正文。
+func einoToolResultBody(content string) string {
+	if strings.HasPrefix(content, einomcp.ToolErrorPrefix) {
+		return strings.TrimPrefix(content, einomcp.ToolErrorPrefix)
+	}
+	return content
+}
+
+// nextAgentEventWithContext 在 ctx 取消时不再无限阻塞于 iter.Next()（工具执行/模型推理期间常见）。
+func nextAgentEventWithContext(ctx context.Context, iter *adk.AsyncIterator[*adk.AgentEvent]) (ev *adk.AgentEvent, ok bool, ctxErr error) {
+	if iter == nil {
+		return nil, false, nil
+	}
+	type nextRes struct {
+		ev *adk.AgentEvent
+		ok bool
+	}
+	ch := make(chan nextRes, 1)
+	go func() {
+		e, o := iter.Next()
+		ch <- nextRes{e, o}
+	}()
+	select {
+	case <-ctx.Done():
+		return nil, false, ctx.Err()
+	case res := <-ch:
+		return res.ev, res.ok, nil
+	}
+}
+
+// recvSchemaMessageStream 消费 ADK Tool 流式结果；ctx 取消时立即返回，避免 amass 等无输出时永久阻塞。
+func recvSchemaMessageStream(ctx context.Context, stream *schema.StreamReader[*schema.Message]) (content, toolCallID string, recvErr error) {
+	if stream == nil {
+		return "", "", nil
+	}
+	type streamMsg struct {
+		chunk *schema.Message
+		err   error
+	}
+	recvCh := make(chan streamMsg, 8)
+	go func() {
+		defer close(recvCh)
+		for {
+			ch, rerr := stream.Recv()
+			recvCh <- streamMsg{chunk: ch, err: rerr}
+			if rerr != nil {
+				return
+			}
+		}
+	}()
+	var buf strings.Builder
+	for {
+		select {
+		case <-ctx.Done():
+			return buf.String(), toolCallID, ctx.Err()
+		case sm, open := <-recvCh:
+			if !open {
+				return buf.String(), toolCallID, nil
+			}
+			rerr := sm.err
+			if errors.Is(rerr, io.EOF) {
+				return buf.String(), toolCallID, nil
+			}
+			if rerr != nil {
+				return buf.String(), toolCallID, rerr
+			}
+			chunk := sm.chunk
+			if chunk == nil {
+				continue
+			}
+			if chunk.Content != "" {
+				buf.WriteString(chunk.Content)
+			}
+			if tid := strings.TrimSpace(chunk.ToolCallID); tid != "" {
+				toolCallID = tid
+			}
+		}
+	}
+}
+
 func buildEinoRunResultFromAccumulated(
 	orchMode string,
 	runAccumulatedMsgs []adk.Message,

internal/multiagent/eino_adk_run_loop_stream_test.go

@@ -0,0 +1,74 @@
+package multiagent
+
+import (
+	"context"
+	"errors"
+	"io"
+	"testing"
+	"time"
+
+	"github.com/cloudwego/eino/schema"
+)
+
+func TestRecvSchemaMessageStream_EOF(t *testing.T) {
+	sr, sw := schema.Pipe[*schema.Message](4)
+	_ = sw.Send(schema.ToolMessage("hello", "tc-1"), nil)
+	sw.Close()
+
+	content, tid, err := recvSchemaMessageStream(context.Background(), sr)
+	if err != nil {
+		t.Fatalf("unexpected err: %v", err)
+	}
+	if content != "hello" {
+		t.Fatalf("content=%q want hello", content)
+	}
+	if tid != "tc-1" {
+		t.Fatalf("toolCallID=%q want tc-1", tid)
+	}
+}
+
+func TestRecvSchemaMessageStream_ContextCancel(t *testing.T) {
+	sr, sw := schema.Pipe[*schema.Message](4)
+	t.Cleanup(func() { sw.Close() })
+
+	ctx, cancel := context.WithCancel(context.Background())
+	go func() {
+		time.Sleep(30 * time.Millisecond)
+		cancel()
+	}()
+
+	content, _, err := recvSchemaMessageStream(ctx, sr)
+	if !errors.Is(err, context.Canceled) {
+		t.Fatalf("want context.Canceled, got %v content=%q", err, content)
+	}
+}
+
+func TestRecvSchemaMessageStream_RecvError(t *testing.T) {
+	sr, sw := schema.Pipe[*schema.Message](4)
+	want := errors.New("stream broken")
+	_ = sw.Send(nil, want)
+	sw.Close()
+
+	_, _, err := recvSchemaMessageStream(context.Background(), sr)
+	if !errors.Is(err, want) {
+		t.Fatalf("want %v, got %v", want, err)
+	}
+}
+
+func TestRecvSchemaMessageStream_NilStream(t *testing.T) {
+	content, tid, err := recvSchemaMessageStream(context.Background(), nil)
+	if err != nil || content != "" || tid != "" {
+		t.Fatalf("nil stream: content=%q tid=%q err=%v", content, tid, err)
+	}
+}
+
+func TestRecvSchemaMessageStream_EOFViaEmptyRead(t *testing.T) {
+	sr, sw := schema.Pipe[*schema.Message](4)
+	_ = sw.Send(nil, io.EOF)
+	sw.Close()
+
+	_, _, err := recvSchemaMessageStream(context.Background(), sr)
+	if err != nil {
+		t.Fatalf("EOF should not surface as error, got %v", err)
+	}
+}

internal/multiagent/eino_chat_model_tail_middleware.go

@@ -0,0 +1,50 @@
+package multiagent
+
+import (
+	"github.com/cloudwego/eino/adk"
+	"go.uber.org/zap"
+)
+
+// einoChatModelTailConfig configures middleware appended after reduction/skill/plantask
+// and immediately before each ChatModel invocation pipeline completes.
+//
+// Order (best practice):
+//  1. system merge — accurate token count for summarization
+//  2. continuation user dedup — drop stale session-resume injections
+//  3. summarization
+//  4. orphan tool prune
+//  5. telemetry
+//  6. model-facing trace snapshot
+type einoChatModelTailConfig struct {
+	logger           *zap.Logger
+	phase            string
+	summarization    adk.ChatModelAgentMiddleware
+	modelName        string
+	conversationID   string
+	trace            *modelFacingTraceHolder
+	skipOrphanPruner bool
+	skipTelemetry    bool
+	skipTrace        bool
+}
+
+func appendEinoChatModelTailMiddlewares(handlers []adk.ChatModelAgentMiddleware, cfg einoChatModelTailConfig) []adk.ChatModelAgentMiddleware {
+	handlers = append(handlers, newSystemMessageNormalizerMiddleware(cfg.logger, cfg.phase))
+	handlers = append(handlers, newContinuationUserDedupMiddleware(cfg.logger, cfg.phase))
+	if cfg.summarization != nil {
+		handlers = append(handlers, cfg.summarization)
+	}
+	if !cfg.skipOrphanPruner {
+		handlers = append(handlers, newOrphanToolPrunerMiddleware(cfg.logger, cfg.phase))
+	}
+	if !cfg.skipTelemetry {
+		if teleMw := newEinoModelInputTelemetryMiddleware(cfg.logger, cfg.modelName, cfg.conversationID, cfg.phase); teleMw != nil {
+			handlers = append(handlers, teleMw)
+		}
+	}
+	if !cfg.skipTrace && cfg.trace != nil {
+		if capMw := newModelFacingTraceMiddleware(cfg.trace); capMw != nil {
+			handlers = append(handlers, capMw)
+		}
+	}
+	return handlers
+}

internal/multiagent/eino_empty_response.go

@@ -0,0 +1,59 @@
+package multiagent
+
+import (
+	"strings"
+	"time"
+
+	"cyberstrike-ai/internal/config"
+)
+
+const defaultEmptyResponseContinueMaxAttempts = 5
+
+// IsEinoEmptyResponseResult 判断 Run 是否以「未捕获助手正文」占位结束（非真实用户可见回复）。
+func IsEinoEmptyResponseResult(result *RunResult) bool {
+	if result == nil {
+		return false
+	}
+	return isEinoEmptyResponseText(result.Response)
+}
+
+func isEinoEmptyResponseText(s string) bool {
+	s = strings.TrimSpace(s)
+	if s == "" {
+		return false
+	}
+	return strings.Contains(s, "no assistant text was captured") ||
+		strings.Contains(s, "未捕获到助手文本输出")
+}
+
+// HasEinoResumeTrace 轨迹非空，续跑才有上下文可恢复。
+func HasEinoResumeTrace(result *RunResult) bool {
+	if result == nil {
+		return false
+	}
+	s := strings.TrimSpace(result.LastAgentTraceInput)
+	return s != "" && s != "[]" && s != "null"
+}
+
+// EmptyResponseContinueMaxAttemptsFromConfig 无助手正文时 Handler 层退避续跑上限；0=默认 5。
+func EmptyResponseContinueMaxAttemptsFromConfig(mw *config.MultiAgentEinoMiddlewareConfig) int {
+	if mw != nil && mw.EmptyResponseContinueMaxAttempts > 0 {
+		return mw.EmptyResponseContinueMaxAttempts
+	}
+	return defaultEmptyResponseContinueMaxAttempts
+}
+
+// EmptyResponseContinueBackoff 与 run_retry 相同指数退避（2s, 4s, 8s… capped）。
+func EmptyResponseContinueBackoff(attempt int, mw *config.MultiAgentEinoMiddlewareConfig) time.Duration {
+	maxBackoff := defaultEinoRunRetryMaxBackoff
+	if mw != nil && mw.RunRetryMaxBackoffSec > 0 {
+		maxBackoff = time.Duration(mw.RunRetryMaxBackoffSec) * time.Second
+	}
+	return einoTransientRetryBackoff(attempt, maxBackoff)
+}
+
+// FormatEmptyResponseContinueUserMessage 系统自动续跑时注入的 user 轮次（不写入 messages 表气泡）。
+func FormatEmptyResponseContinueUserMessage() string {
+	return strings.TrimSpace(`【系统自动续跑 / Auto resume】
+上一轮 Eino 会话未产出可见助手正文（可能流式中断或仅完成工具调用）。请基于已有轨迹与工具结果继续推进，并给出阶段性总结；勿重复已完成步骤。`)
+}

internal/multiagent/eino_empty_response_test.go

@@ -2,20 +2,37 @@ package multiagent
 
 import "testing"
 
-func TestShouldEinoEmptyResponseContinue(t *testing.T) {
-	t.Parallel()
-	hint := "(empty hint)"
-	out := &RunResult{Response: hint}
-	if !shouldEinoEmptyResponseContinue(out, hint, 3, 1) {
-		t.Fatal("expected continue when response is empty hint and trace grew")
+func TestIsEinoEmptyResponseResult(t *testing.T) {
+	empty := &RunResult{
+		Response: "(Eino ADK single-agent session completed but no assistant text was captured. Check process details or logs.) " +
+			"（Eino ADK 单代理会话已完成，但未捕获到助手文本输出。请查看过程详情或日志。）",
 	}
-	if shouldEinoEmptyResponseContinue(out, hint, 1, 1) {
-		t.Fatal("expected no continue when trace did not grow")
+	if !IsEinoEmptyResponseResult(empty) {
+		t.Fatal("expected empty placeholder response")
 	}
-	if shouldEinoEmptyResponseContinue(&RunResult{Response: "hello"}, hint, 3, 1) {
-		t.Fatal("expected no continue when response has content")
+	ok := &RunResult{Response: "扫描完成，发现 2 个开放端口。"}
+	if IsEinoEmptyResponseResult(ok) {
+		t.Fatalf("expected real response, got placeholder match")
 	}
-	if shouldEinoEmptyResponseContinue(nil, hint, 3, 1) {
-		t.Fatal("expected no continue for nil result")
+	if IsEinoEmptyResponseResult(nil) {
+		t.Fatal("nil result should be false")
+	}
+}
+
+func TestHasEinoResumeTrace(t *testing.T) {
+	if HasEinoResumeTrace(nil) {
+		t.Fatal("nil")
+	}
+	if HasEinoResumeTrace(&RunResult{LastAgentTraceInput: "[]"}) {
+		t.Fatal("enable resume on empty trace")
+	}
+	if !HasEinoResumeTrace(&RunResult{LastAgentTraceInput: `[{"role":"user","content":"hi"}]`}) {
+		t.Fatal("expected resume trace")
+	}
+}
+
+func TestEmptyResponseContinueMaxAttemptsFromConfig(t *testing.T) {
+	if got := EmptyResponseContinueMaxAttemptsFromConfig(nil); got != defaultEmptyResponseContinueMaxAttempts {
+		t.Fatalf("default: got %d want %d", got, defaultEmptyResponseContinueMaxAttempts)
 	}
 }

internal/multiagent/eino_execute_failure_format_test.go

@@ -0,0 +1,114 @@
+package multiagent
+
+import (
+	"context"
+	"errors"
+	"io"
+	"strings"
+	"testing"
+
+	"cyberstrike-ai/internal/einomcp"
+	"cyberstrike-ai/internal/security"
+
+	"github.com/cloudwego/eino/adk/filesystem"
+	"github.com/cloudwego/eino/schema"
+)
+
+type mockStreamingShellExitFail struct {
+	output string
+	code   int
+}
+
+func (m *mockStreamingShellExitFail) ExecuteStreaming(ctx context.Context, input *filesystem.ExecuteRequest) (*schema.StreamReader[*filesystem.ExecuteResponse], error) {
+	outR, outW := schema.Pipe[*filesystem.ExecuteResponse](4)
+	go func() {
+		defer outW.Close()
+		if m.output != "" {
+			_ = outW.Send(&filesystem.ExecuteResponse{Output: m.output}, nil)
+		}
+		code := m.code
+		_ = outW.Send(&filesystem.ExecuteResponse{ExitCode: &code}, nil)
+	}()
+	return outR, nil
+}
+
+func TestEinoStreamingShellWrap_CommandFailureFormat(t *testing.T) {
+	inner := &mockStreamingShellExitFail{
+		output: "sudo: a password is required\n",
+		code:   1,
+	}
+	notify := einomcp.NewToolInvokeNotifyHolder()
+	var firedBody string
+	var firedSuccess bool
+	var firedErr error
+	notify.Set(func(toolCallID, toolName, einoAgent string, success bool, content string, invokeErr error) {
+		firedBody = content
+		firedSuccess = success
+		firedErr = invokeErr
+	})
+	wrap := &einoStreamingShellWrap{inner: inner, invokeNotify: notify}
+	sr, err := wrap.ExecuteStreaming(context.Background(), &filesystem.ExecuteRequest{Command: "sudo whoami"})
+	if err != nil {
+		t.Fatalf("ExecuteStreaming: %v", err)
+	}
+	defer sr.Close()
+
+	var stream strings.Builder
+	for {
+		resp, rerr := sr.Recv()
+		if errors.Is(rerr, io.EOF) {
+			break
+		}
+		if rerr != nil {
+			t.Fatalf("recv: %v", rerr)
+		}
+		if resp != nil {
+			stream.WriteString(resp.Output)
+		}
+	}
+
+	if firedSuccess {
+		t.Fatal("expected success=false")
+	}
+	var exitErr *ExecuteExitError
+	if !errors.As(firedErr, &exitErr) || exitErr.Code != 1 {
+		t.Fatalf("expected ExecuteExitError code 1, got %v", firedErr)
+	}
+	if !strings.HasPrefix(firedBody, einomcp.ToolErrorPrefix) {
+		t.Fatalf("missing tool error prefix: %q", firedBody)
+	}
+	body := strings.TrimPrefix(firedBody, einomcp.ToolErrorPrefix)
+	if body != security.FormatCommandFailureResult(1, "sudo: a password is required\n") {
+		t.Fatalf("fire body = %q", body)
+	}
+	if !strings.Contains(stream.String(), "sudo:") {
+		t.Fatalf("stream missing sudo output: %q", stream.String())
+	}
+	if strings.Contains(stream.String(), "command exited with non-zero") {
+		t.Fatalf("stream has legacy noise: %q", stream.String())
+	}
+	if strings.Contains(stream.String(), "执行未正常结束") {
+		t.Fatalf("stream has abnormal tail: %q", stream.String())
+	}
+	if !security.IsCommandFailureResult(stream.String()) {
+		t.Fatalf("stream missing failure status line: %q", stream.String())
+	}
+	if tail := friendlyEinoExecuteInvokeTail(firedErr); tail != "" {
+		t.Fatalf("unexpected invoke tail: %q", tail)
+	}
+	if !einoToolResultIsError("execute", firedBody) {
+		t.Fatal("expected isError for execute failure")
+	}
+}
+
+func TestFriendlyEinoExecuteInvokeTail(t *testing.T) {
+	if friendlyEinoExecuteInvokeTail(&ExecuteExitError{Code: 1}) != "" {
+		t.Fatal("exit error should not get abnormal tail")
+	}
+	if !strings.Contains(friendlyEinoExecuteInvokeTail(context.DeadlineExceeded), "Timed out") {
+		t.Fatal("deadline should get timeout hint")
+	}
+	if friendlyEinoExecuteInvokeTail(errors.New("broken pipe")) == "" {
+		t.Fatal("unexpected error should get tail")
+	}
+}

internal/multiagent/eino_execute_monitor.go

@@ -7,11 +7,25 @@ import (
 	"cyberstrike-ai/internal/einomcp"
 )
 
-// newEinoExecuteMonitorCallback 在 Eino filesystem execute 结束时写入 MCP 监控库并 recorder(executionId)，
-// 与 CallTool 路径一致，供助手消息展示「渗透测试详情」芯片。
-func newEinoExecuteMonitorCallback(ag *agent.Agent, recorder einomcp.ExecutionRecorder) func(toolCallID, command, stdout string, success bool, invokeErr error) {
-	return func(toolCallID, command, stdout string, success bool, invokeErr error) {
-		if ag == nil || recorder == nil {
+// newEinoExecuteMonitorCallbacks 在 Eino filesystem execute 开始/结束时写入 MCP 监控库并 recorder(executionId)，
+// 与 CallTool 路径一致，使监控页能展示「执行中」状态。
+func newEinoExecuteMonitorCallbacks(ag *agent.Agent, recorder einomcp.ExecutionRecorder) (
+	begin func(toolCallID, command string) string,
+	finish func(executionID, toolCallID, command, stdout string, success bool, invokeErr error),
+) {
+	begin = func(toolCallID, command string) string {
+		if ag == nil {
+			return ""
+		}
+		args := map[string]interface{}{"command": command}
+		id := ag.BeginLocalToolExecution("execute", args)
+		if id != "" && recorder != nil {
+			recorder(id, toolCallID)
+		}
+		return id
+	}
+	finish = func(executionID, toolCallID, command, stdout string, success bool, invokeErr error) {
+		if ag == nil {
 			return
 		}
 		var err error
@@ -23,9 +37,10 @@ func newEinoExecuteMonitorCallback(ag *agent.Agent, recorder einomcp.ExecutionRe
 			}
 		}
 		args := map[string]interface{}{"command": command}
-		id := ag.RecordLocalToolExecution("execute", args, stdout, err)
-		if id != "" {
+		id := ag.FinishLocalToolExecution(executionID, "execute", args, stdout, err)
+		if id != "" && recorder != nil && executionID == "" {
 			recorder(id, toolCallID)
 		}
 	}
+	return begin, finish
 }

internal/multiagent/eino_execute_streaming_wrap.go

@@ -6,9 +6,11 @@ import (
 	"fmt"
 	"io"
 	"strings"
+	"sync"
 	"time"
 
 	"cyberstrike-ai/internal/einomcp"
+	"cyberstrike-ai/internal/mcp"
 	"cyberstrike-ai/internal/security"
 
 	"github.com/cloudwego/eino/adk/filesystem"
@@ -49,7 +51,7 @@ func einoExecuteRecvErrIsToolTimeout(rerr error, tctx context.Context) bool {
 // 对「完全后台」命令自动开启 RunInBackendGround，与 local.runCmdInBackground 行为对齐。
 //
 // 使用 Pipe 将内层流转发给调用方：在 inner EOF 后、关闭 Pipe 前同步调用 ToolInvokeNotify.Fire，
-// 保证 run loop 在模型开始下一轮输出前已记录 execute 结果（用于 UI 与「重复助手复述」去重）。
+// run loop 收到 Fire 后立即推送 tool_result（toolResultSent 去重），避免 ADK Tool 事件迟到时 UI 卡在「执行中」。
 //
 // 若 inner 在校验阶段直接返回 error（未建立 reader），不会进入下方 goroutine，也必须 Fire；
 // 否则 pending tool_call 要等整轮 run 结束才被 force-close，与已展示的助手/工具软错误文案不同步。
@@ -61,8 +63,11 @@ type einoStreamingShellWrap struct {
 	outputChunk func(toolName, toolCallID, chunk string)
 	// toolTimeoutMinutes 与 agent.tool_timeout_minutes 对齐；>0 时对单次 execute 套用 context 超时（与 MCP 工具经 executeToolViaMCP 行为一致）。0 表示仅依赖上层 ctx（如整任务 10h 上限）。
 	toolTimeoutMinutes int
-	// recordMonitor 在 execute 流结束后写入 tool_executions 并 recorder(executionId)，使「渗透测试详情」与常规 MCP 一致。
-	recordMonitor func(toolCallID, command, stdout string, success bool, invokeErr error)
+	// shellNoOutputTimeoutSec：无任何输出时的空闲秒数；0=关闭。
+	shellNoOutputTimeoutSec int
+	// beginMonitor 在 execute 开始时写入 running 状态；finishMonitor 在流结束后更新为 completed/failed。
+	beginMonitor func(toolCallID, command string) string
+	finishMonitor func(executionID, toolCallID, command, stdout string, success bool, invokeErr error)
 }
 
 func (w *einoStreamingShellWrap) ExecuteStreaming(ctx context.Context, input *filesystem.ExecuteRequest) (*schema.StreamReader[*filesystem.ExecuteResponse], error) {
@@ -74,43 +79,65 @@ func (w *einoStreamingShellWrap) ExecuteStreaming(ctx context.Context, input *fi
 	}
 	req := *input
 	userCmd := strings.TrimSpace(req.Command)
+	tid := strings.TrimSpace(compose.GetToolCallID(ctx))
+	agentTag := strings.TrimSpace(w.einoAgentName)
 	if security.IsBackgroundShellCommand(req.Command) && !req.RunInBackendGround {
 		req.RunInBackendGround = true
 	}
 	req.Command = prependPythonUnbufferedEnv(req.Command)
-	tid := strings.TrimSpace(compose.GetToolCallID(ctx))
-	agentTag := strings.TrimSpace(w.einoAgentName)
+	convID := mcp.MCPConversationIDFromContext(ctx)
+	execReg := mcp.EinoExecuteRunRegistryFromContext(ctx)
 
-	execCtx := ctx
-	var execCancel context.CancelFunc
+	var monitorExecID string
+	if w.beginMonitor != nil {
+		monitorExecID = w.beginMonitor(tid, userCmd)
+	}
+	if monitorExecID != "" && convID != "" {
+		if toolReg := mcp.ToolRunRegistryFromContext(ctx); toolReg != nil {
+			toolReg.RegisterRunningTool(convID, monitorExecID)
+		}
+	}
+	toolRunReg := mcp.ToolRunRegistryFromContext(ctx)
+
+	execCtx, execCancel := context.WithCancel(ctx)
+	var timeoutCancel context.CancelFunc
 	if w.toolTimeoutMinutes > 0 {
-		execCtx, execCancel = context.WithTimeout(ctx, time.Duration(w.toolTimeoutMinutes)*time.Minute)
+		execCtx, timeoutCancel = context.WithTimeout(execCtx, time.Duration(w.toolTimeoutMinutes)*time.Minute)
+	}
+	if execReg != nil && convID != "" {
+		execReg.RegisterActiveEinoExecute(convID, execCancel)
 	}
 
 	sr, err := w.inner.ExecuteStreaming(execCtx, &req)
 	if err != nil {
+		if timeoutCancel != nil {
+			timeoutCancel()
+		}
 		if execCancel != nil {
 			execCancel()
 		}
 		if einoExecuteRecvErrIsToolTimeout(err, execCtx) {
 			hint := "\n\n" + einoExecuteTimeoutUserHint() + "\n"
-			if w.recordMonitor != nil {
-				w.recordMonitor(tid, userCmd, hint, false, context.DeadlineExceeded)
+			if w.finishMonitor != nil {
+				w.finishMonitor(monitorExecID, tid, userCmd, hint, false, context.DeadlineExceeded)
 			}
 			if w.invokeNotify != nil && tid != "" {
 				w.invokeNotify.Fire(tid, "execute", agentTag, false, hint, context.DeadlineExceeded)
 			}
 			return schema.StreamReaderFromArray([]*filesystem.ExecuteResponse{{Output: hint}}), nil
 		}
-		if w.recordMonitor != nil {
-			w.recordMonitor(tid, userCmd, "", false, err)
+		if w.finishMonitor != nil {
+			w.finishMonitor(monitorExecID, tid, userCmd, "", false, err)
 		}
 		if w.invokeNotify != nil && tid != "" {
 			w.invokeNotify.Fire(tid, "execute", agentTag, false, "", err)
 		}
 		return nil, err
 	}
-	if sr == nil || w.invokeNotify == nil {
+	if sr == nil {
+		if timeoutCancel != nil {
+			timeoutCancel()
+		}
 		if execCancel != nil {
 			execCancel()
 		}
@@ -119,11 +146,35 @@ func (w *einoStreamingShellWrap) ExecuteStreaming(ctx context.Context, input *fi
 
 	outR, outW := schema.Pipe[*filesystem.ExecuteResponse](32)
 
-	go func(inner *schema.StreamReader[*filesystem.ExecuteResponse], command string, cancel context.CancelFunc, tctx context.Context) {
-		defer inner.Close()
+	go func(inner *schema.StreamReader[*filesystem.ExecuteResponse], command string, cancel context.CancelFunc, timeoutCleanup context.CancelFunc, tctx context.Context, conversationID string, reg mcp.EinoExecuteRunRegistry, toolReg mcp.ToolRunRegistry, execID string, toolCallID string, noOutputSec int) {
+		var innerCloseOnce sync.Once
+		closeInner := func() {
+			innerCloseOnce.Do(func() { inner.Close() })
+		}
+		defer closeInner()
+		if timeoutCleanup != nil {
+			defer timeoutCleanup()
+		}
 		if cancel != nil {
 			defer cancel()
 		}
+		if reg != nil && conversationID != "" {
+			defer reg.UnregisterActiveEinoExecute(conversationID)
+		}
+		if toolReg != nil && conversationID != "" && execID != "" {
+			defer toolReg.UnregisterRunningTool(conversationID, execID)
+		}
+
+		// ctx 取消时关闭内层流，避免 amass 等长时间无换行输出时 Recv 永久阻塞。
+		stopWatch := make(chan struct{})
+		go func() {
+			select {
+			case <-tctx.Done():
+				closeInner()
+			case <-stopWatch:
+			}
+		}()
+		defer close(stopWatch)
 
 		var sb strings.Builder
 		success := true
@@ -131,46 +182,103 @@ func (w *einoStreamingShellWrap) ExecuteStreaming(ctx context.Context, input *fi
 		exitCode := 0
 		hasExitCode := false
 
+		idleWatch := security.NewShellInactivityWatch(noOutputSec)
+		if idleWatch != nil {
+			defer idleWatch.Stop()
+		}
+
+		type execRecvMsg struct {
+			resp *filesystem.ExecuteResponse
+			err  error
+		}
+		recvCh := make(chan execRecvMsg, 1)
+		go func() {
+			for {
+				resp, rerr := inner.Recv()
+				recvCh <- execRecvMsg{resp: resp, err: rerr}
+				if rerr != nil {
+					return
+				}
+			}
+		}()
+
+		fireInactivityTimeout := func() {
+			success = false
+			invokeErr = fmt.Errorf("shell inactivity timeout (%ds)", idleWatch.Sec)
+			msg := security.ShellNoOutputTimeoutMessage(idleWatch.Sec)
+			_ = outW.Send(&filesystem.ExecuteResponse{Output: msg}, nil)
+			sb.WriteString(msg)
+			if w.outputChunk != nil && toolCallID != "" {
+				w.outputChunk("execute", toolCallID, msg)
+			}
+			if cancel != nil {
+				cancel()
+			}
+			closeInner()
+		}
+
+	recvLoop:
 		for {
-			resp, rerr := inner.Recv()
-			if errors.Is(rerr, io.EOF) {
-				break
+			var idleCh <-chan struct{}
+			if idleWatch != nil {
+				idleCh = idleWatch.Expired
 			}
-			if rerr != nil {
-				success = false
-				invokeErr = rerr
-				// 单次 execute 超时须与 MCP 工具一致：写入工具结果尾标、继续迭代，不得向 ADK 流注入硬错误。
-				if einoExecuteRecvErrIsToolTimeout(rerr, tctx) {
-					invokeErr = context.DeadlineExceeded
-					break
+			select {
+			case <-idleCh:
+				fireInactivityTimeout()
+				break recvLoop
+			case msg := <-recvCh:
+				rerr := msg.err
+				resp := msg.resp
+				if errors.Is(rerr, io.EOF) {
+					break recvLoop
 				}
-				_ = outW.Send(nil, rerr)
-				break
-			}
-			if resp != nil {
-				if resp.ExitCode != nil {
-					hasExitCode = true
-					exitCode = *resp.ExitCode
-				}
-				var appended string
-				if resp.Output != "" {
-					sb.WriteString(resp.Output)
-					appended = resp.Output
-				}
-				if w.outputChunk != nil && strings.TrimSpace(appended) != "" {
-					w.outputChunk("execute", tid, appended)
-				}
-				if outW.Send(resp, nil) {
+				if rerr != nil {
 					success = false
-					invokeErr = fmt.Errorf("execute stream closed by consumer")
-					break
+					invokeErr = rerr
+					if einoExecuteRecvErrIsToolTimeout(rerr, tctx) {
+						invokeErr = context.DeadlineExceeded
+						break recvLoop
+					}
+					if errors.Is(rerr, context.Canceled) || (tctx != nil && errors.Is(tctx.Err(), context.Canceled)) {
+						invokeErr = context.Canceled
+						break recvLoop
+					}
+					_ = outW.Send(nil, rerr)
+					break recvLoop
+				}
+				if resp != nil {
+					if resp.ExitCode != nil {
+						hasExitCode = true
+						exitCode = *resp.ExitCode
+						continue
+					}
+					var appended string
+					if resp.Output != "" {
+						if security.IsLegacyShellExitNoise(resp.Output) {
+							continue
+						}
+						if idleWatch != nil {
+							idleWatch.Bump()
+						}
+						sb.WriteString(resp.Output)
+						appended = resp.Output
+					}
+					if w.outputChunk != nil && strings.TrimSpace(appended) != "" {
+						w.outputChunk("execute", toolCallID, appended)
+					}
+					if outW.Send(resp, nil) {
+						success = false
+						invokeErr = fmt.Errorf("execute stream closed by consumer")
+						break recvLoop
+					}
 				}
 			}
 		}
 
 		if success && hasExitCode && exitCode != 0 {
 			success = false
-			invokeErr = fmt.Errorf("execute exited with code %d", exitCode)
+			invokeErr = &ExecuteExitError{Code: exitCode}
 		}
 		// WithTimeout 触发后，子进程常被信号结束，local 侧多报 exit -1 / canceled，错误链里不一定带 DeadlineExceeded。
 		// 用执行所用 ctx 归一化，便于 UI 展示「超时」而非含糊的 -1。
@@ -178,6 +286,21 @@ func (w *einoStreamingShellWrap) ExecuteStreaming(ctx context.Context, input *fi
 			success = false
 			invokeErr = context.DeadlineExceeded
 		}
+		// 用户「中断并继续」终止 execute：合并说明进工具结果（与 MCP CancelToolExecutionWithNote 一致）。
+		partialStreamed := sb.String()
+		var abortNote string
+		if reg != nil && conversationID != "" && (invokeErr != nil || errors.Is(tctx.Err(), context.Canceled)) {
+			if note := reg.TakeEinoExecuteAbortNote(conversationID); note != "" {
+				abortNote = note
+				merged := mcp.MergePartialToolOutputAndAbortNote(partialStreamed, note)
+				sb.Reset()
+				sb.WriteString(merged)
+				if invokeErr == nil {
+					success = false
+					invokeErr = context.Canceled
+				}
+			}
+		}
 		// ADK 从本 Pipe 拼出 tool 消息正文；仅 Notify 尾标不会进入模型上下文。超时句写入流，与 UI 一致。
 		if invokeErr != nil && errors.Is(invokeErr, context.DeadlineExceeded) {
 			hint := "\n\n" + einoExecuteTimeoutUserHint() + "\n"
@@ -187,12 +310,32 @@ func (w *einoStreamingShellWrap) ExecuteStreaming(ctx context.Context, input *fi
 			}
 			sb.WriteString(hint)
 		}
-		if w.recordMonitor != nil {
-			w.recordMonitor(tid, command, sb.String(), success, invokeErr)
+		// 中断时循环内已逐行写入 stdout；此处只追加 USER INTERRUPT NOTE，避免整段输出重复。
+		if invokeErr != nil && errors.Is(invokeErr, context.Canceled) && abortNote != "" {
+			if partialStreamed != "" {
+				_ = outW.Send(&filesystem.ExecuteResponse{Output: "\n\n" + mcp.AbortNoteBannerForModel + "\n" + abortNote}, nil)
+			} else if text := strings.TrimSpace(sb.String()); text != "" {
+				_ = outW.Send(&filesystem.ExecuteResponse{Output: text + "\n"}, nil)
+			}
+		}
+		rawOutput := sb.String()
+		fireBody := rawOutput
+		if !success && hasExitCode && exitCode != 0 {
+			statusLine := security.ExecuteFailureStatusLine(exitCode)
+			if !strings.Contains(rawOutput, "命令执行失败:") {
+				_ = outW.Send(&filesystem.ExecuteResponse{Output: statusLine}, nil)
+				sb.WriteString(statusLine)
+			}
+			fireBody = einomcp.ToolErrorPrefix + security.FormatCommandFailureResult(exitCode, rawOutput)
+		}
+		if w.finishMonitor != nil {
+			w.finishMonitor(execID, toolCallID, command, sb.String(), success, invokeErr)
+		}
+		if w.invokeNotify != nil {
+			w.invokeNotify.Fire(toolCallID, "execute", agentTag, success, fireBody, invokeErr)
 		}
-		w.invokeNotify.Fire(tid, "execute", agentTag, success, sb.String(), invokeErr)
 		outW.Close()
-	}(sr, userCmd, execCancel, execCtx)
+	}(sr, userCmd, execCancel, timeoutCancel, execCtx, convID, execReg, toolRunReg, monitorExecID, tid, w.shellNoOutputTimeoutSec)
 
 	return outR, nil
 }

internal/multiagent/eino_execute_streaming_wrap_test.go

@@ -9,6 +9,7 @@ import (
 	"time"
 
 	"cyberstrike-ai/internal/einomcp"
+	"cyberstrike-ai/internal/mcp"
 
 	"github.com/cloudwego/eino/adk/filesystem"
 	"github.com/cloudwego/eino/schema"
@@ -18,9 +19,15 @@ type mockStreamingShell struct {
 	immediateErr error
 	recvErr      error
 	output       string
+	called       bool
+	lastCommand  string
 }
 
 func (m *mockStreamingShell) ExecuteStreaming(ctx context.Context, input *filesystem.ExecuteRequest) (*schema.StreamReader[*filesystem.ExecuteResponse], error) {
+	m.called = true
+	if input != nil {
+		m.lastCommand = input.Command
+	}
 	if m.immediateErr != nil {
 		return nil, m.immediateErr
 	}
@@ -37,6 +44,129 @@ func (m *mockStreamingShell) ExecuteStreaming(ctx context.Context, input *filesy
 	return outR, nil
 }
 
+func TestEinoStreamingShellWrap_PreparesNonInteractiveCommand(t *testing.T) {
+	inner := &mockStreamingShell{output: "ok\n"}
+	wrap := &einoStreamingShellWrap{inner: inner}
+	sr, err := wrap.ExecuteStreaming(context.Background(), &filesystem.ExecuteRequest{Command: "echo ok"})
+	if err != nil {
+		t.Fatalf("ExecuteStreaming: %v", err)
+	}
+	defer sr.Close()
+	for {
+		_, rerr := sr.Recv()
+		if errors.Is(rerr, io.EOF) {
+			break
+		}
+		if rerr != nil {
+			t.Fatalf("recv: %v", rerr)
+		}
+	}
+	if !strings.Contains(inner.lastCommand, "PYTHONUNBUFFERED=1") {
+		t.Fatalf("missing python unbuffer in inner command: %q", inner.lastCommand)
+	}
+}
+
+func TestEinoStreamingShellWrap_NoOutputTimeout(t *testing.T) {
+	inner := &mockStreamingShellHanging{}
+	notify := einomcp.NewToolInvokeNotifyHolder()
+	var fired string
+	notify.Set(func(toolCallID, toolName, einoAgent string, success bool, content string, invokeErr error) {
+		fired = content
+	})
+	wrap := &einoStreamingShellWrap{
+		inner:                   inner,
+		invokeNotify:            notify,
+		shellNoOutputTimeoutSec: 1,
+	}
+	sr, err := wrap.ExecuteStreaming(context.Background(), &filesystem.ExecuteRequest{Command: "sudo whoami"})
+	if err != nil {
+		t.Fatalf("ExecuteStreaming: %v", err)
+	}
+	defer sr.Close()
+	var got strings.Builder
+	for {
+		resp, rerr := sr.Recv()
+		if errors.Is(rerr, io.EOF) {
+			break
+		}
+		if rerr != nil {
+			t.Fatalf("recv: %v", rerr)
+		}
+		if resp != nil {
+			got.WriteString(resp.Output)
+		}
+	}
+	if !inner.called {
+		t.Fatal("inner shell should run (no command blacklist)")
+	}
+	out := got.String()
+	if !strings.Contains(out, "没有新的输出") && !strings.Contains(out, "no new output") {
+		t.Fatalf("expected inactivity timeout message, got: %q notify=%q", out, fired)
+	}
+}
+
+type mockStreamingShellPartialThenHang struct {
+	called bool
+}
+
+func (m *mockStreamingShellPartialThenHang) ExecuteStreaming(ctx context.Context, input *filesystem.ExecuteRequest) (*schema.StreamReader[*filesystem.ExecuteResponse], error) {
+	m.called = true
+	outR, outW := schema.Pipe[*filesystem.ExecuteResponse](4)
+	go func() {
+		_ = outW.Send(&filesystem.ExecuteResponse{Output: "[sudo] password:\n"}, nil)
+		<-ctx.Done()
+		outW.Close()
+	}()
+	return outR, nil
+}
+
+func TestEinoStreamingShellWrap_InactivityAfterPartialOutput(t *testing.T) {
+	inner := &mockStreamingShellPartialThenHang{}
+	wrap := &einoStreamingShellWrap{
+		inner:                   inner,
+		shellNoOutputTimeoutSec: 1,
+	}
+	start := time.Now()
+	sr, err := wrap.ExecuteStreaming(context.Background(), &filesystem.ExecuteRequest{Command: "sudo whoami"})
+	if err != nil {
+		t.Fatalf("ExecuteStreaming: %v", err)
+	}
+	defer sr.Close()
+	var got strings.Builder
+	for {
+		resp, rerr := sr.Recv()
+		if errors.Is(rerr, io.EOF) {
+			break
+		}
+		if rerr != nil {
+			t.Fatalf("recv: %v", rerr)
+		}
+		if resp != nil {
+			got.WriteString(resp.Output)
+		}
+	}
+	if time.Since(start) > 5*time.Second {
+		t.Fatalf("expected inactivity timeout ~1s, took %v", time.Since(start))
+	}
+	if !strings.Contains(got.String(), "没有新的输出") && !strings.Contains(got.String(), "no new output") {
+		t.Fatalf("expected inactivity message, got: %q", got.String())
+	}
+}
+
+type mockStreamingShellHanging struct {
+	called bool
+}
+
+func (m *mockStreamingShellHanging) ExecuteStreaming(ctx context.Context, input *filesystem.ExecuteRequest) (*schema.StreamReader[*filesystem.ExecuteResponse], error) {
+	m.called = true
+	outR, outW := schema.Pipe[*filesystem.ExecuteResponse](4)
+	go func() {
+		<-ctx.Done()
+		outW.Close()
+	}()
+	return outR, nil
+}
+
 func TestEinoExecuteRecvErrIsToolTimeout(t *testing.T) {
 	tctx, cancel := context.WithTimeout(context.Background(), time.Millisecond)
 	defer cancel()
@@ -122,6 +252,94 @@ func TestEinoStreamingShellWrap_ToolTimeoutRecvErrIsSoft(t *testing.T) {
 	}
 }
 
+func TestEinoStreamingShellWrap_CapturesOutputWithToolTimeout(t *testing.T) {
+	inner := &mockStreamingShell{output: "100\n"}
+	notify := einomcp.NewToolInvokeNotifyHolder()
+	var firedContent string
+	notify.Set(func(toolCallID, toolName, einoAgent string, success bool, content string, invokeErr error) {
+		firedContent = content
+	})
+	wrap := &einoStreamingShellWrap{
+		inner:              inner,
+		invokeNotify:       notify,
+		toolTimeoutMinutes: 60,
+	}
+	sr, err := wrap.ExecuteStreaming(context.Background(), &filesystem.ExecuteRequest{Command: "echo 100"})
+	if err != nil {
+		t.Fatalf("ExecuteStreaming: %v", err)
+	}
+	defer sr.Close()
+
+	var got strings.Builder
+	for {
+		resp, rerr := sr.Recv()
+		if errors.Is(rerr, io.EOF) {
+			break
+		}
+		if rerr != nil {
+			t.Fatalf("unexpected stream error: %v", rerr)
+		}
+		if resp != nil && resp.Output != "" {
+			got.WriteString(resp.Output)
+		}
+	}
+	if !strings.Contains(got.String(), "100") {
+		t.Fatalf("stream output = %q, want contains 100", got.String())
+	}
+	if !strings.Contains(firedContent, "100") {
+		t.Fatalf("notify content = %q, want contains 100", firedContent)
+	}
+}
+
+func TestEinoStreamingShellWrap_AbortNoteDoesNotDuplicateStreamedOutput(t *testing.T) {
+	inner := &mockStreamingShell{output: "line1\nline2\n", recvErr: context.Canceled}
+	notify := einomcp.NewToolInvokeNotifyHolder()
+	wrap := &einoStreamingShellWrap{
+		inner:        inner,
+		invokeNotify: notify,
+	}
+	reg := &abortNoteTestRegistry{note: "改成20次"}
+	ctx := mcp.WithEinoExecuteRunRegistry(
+		mcp.WithMCPConversationID(context.Background(), "conv-abort-dup"),
+		reg,
+	)
+	sr, err := wrap.ExecuteStreaming(ctx, &filesystem.ExecuteRequest{Command: "ping -c 10 baidu.com"})
+	if err != nil {
+		t.Fatalf("ExecuteStreaming: %v", err)
+	}
+	defer sr.Close()
+
+	var got strings.Builder
+	for {
+		resp, rerr := sr.Recv()
+		if errors.Is(rerr, io.EOF) {
+			break
+		}
+		if rerr != nil {
+			t.Fatalf("unexpected stream error: %v", rerr)
+		}
+		if resp != nil && resp.Output != "" {
+			got.WriteString(resp.Output)
+		}
+	}
+	out := got.String()
+	if strings.Count(out, "line1") != 1 || strings.Count(out, "line2") != 1 {
+		t.Fatalf("stream duplicated stdout: %q", out)
+	}
+	if !strings.Contains(out, "改成20次") {
+		t.Fatalf("stream missing abort note: %q", out)
+	}
+}
+
+type abortNoteTestRegistry struct {
+	note string
+}
+
+func (r *abortNoteTestRegistry) RegisterActiveEinoExecute(string, context.CancelFunc) {}
+func (r *abortNoteTestRegistry) UnregisterActiveEinoExecute(string)                   {}
+func (r *abortNoteTestRegistry) AbortActiveEinoExecute(string, string) bool           { return false }
+func (r *abortNoteTestRegistry) TakeEinoExecuteAbortNote(string) string               { return r.note }
+
 func TestEinoStreamingShellWrap_NonTimeoutRecvErrStillHard(t *testing.T) {
 	inner := &mockStreamingShell{recvErr: errors.New("broken pipe")}
 	wrap := &einoStreamingShellWrap{inner: inner}

internal/multiagent/eino_filesystem_tool_monitor.go

@@ -63,10 +63,43 @@ func toolCallArgsFromAccumulated(msgs []adk.Message, toolCallID, expectToolName
 	return map[string]interface{}{}
 }
 
+// beginEinoADKFilesystemToolMonitor 在 Eino ADK filesystem 工具开始调用时写入 running 状态。
+func beginEinoADKFilesystemToolMonitor(
+	ag *agent.Agent,
+	rec einomcp.ExecutionRecorder,
+	binder *MCPExecutionBinder,
+	toolCallID, toolName string,
+) {
+	if ag == nil || rec == nil {
+		return
+	}
+	name := strings.TrimSpace(toolName)
+	if name == "" || strings.EqualFold(name, "execute") {
+		return
+	}
+	if !isBuiltinEinoADKFilesystemToolName(name) {
+		return
+	}
+	tid := strings.TrimSpace(toolCallID)
+	if tid == "" {
+		return
+	}
+	storedName := "eino_fs::" + strings.ToLower(name)
+	id := ag.BeginLocalToolExecution(storedName, map[string]interface{}{})
+	if id == "" {
+		return
+	}
+	rec(id, tid)
+	if binder != nil {
+		binder.Bind(tid, id)
+	}
+}
+
 // recordEinoADKFilesystemToolMonitor 将 Eino ADK filesystem 中间件工具结果写入 MCP 监控（与 execute / MCP 桥芯片一致）。
 func recordEinoADKFilesystemToolMonitor(
 	ag *agent.Agent,
 	rec einomcp.ExecutionRecorder,
+	binder *MCPExecutionBinder,
 	toolName string,
 	toolCallID string,
 	msgs []adk.Message,
@@ -94,8 +127,12 @@ func recordEinoADKFilesystemToolMonitor(
 			invErr = errors.New(t)
 		}
 	}
-	id := ag.RecordLocalToolExecution(storedName, args, resultText, invErr)
-	if id != "" {
+	execID := ""
+	if binder != nil {
+		execID = binder.ExecutionID(toolCallID)
+	}
+	id := ag.FinishLocalToolExecution(execID, storedName, args, resultText, invErr)
+	if id != "" && execID == "" {
 		rec(id, toolCallID)
 	}
 }

internal/multiagent/eino_middleware.go

@@ -243,17 +243,14 @@ func prependEinoMiddlewares(
 	return outTools, extraHandlers, toolSearchActive, nil
 }
 
-func deepExtrasFromConfig(ma *config.MultiAgentConfig) (outputKey string, retry *adk.ModelRetryConfig, taskDesc func(context.Context, []adk.Agent) (string, error)) {
+func deepExtrasFromConfig(ma *config.MultiAgentConfig) (outputKey string, taskDesc func(context.Context, []adk.Agent) (string, error)) {
 	if ma == nil {
-		return "", nil, nil
+		return "", nil
 	}
 	mw := ma.EinoMiddleware
 	if k := strings.TrimSpace(mw.DeepOutputKey); k != "" {
 		outputKey = k
 	}
-	if mw.DeepModelRetryMaxRetries > 0 {
-		retry = &adk.ModelRetryConfig{MaxRetries: mw.DeepModelRetryMaxRetries}
-	}
 	prefix := strings.TrimSpace(mw.TaskToolDescriptionPrefix)
 	if prefix != "" {
 		taskDesc = func(ctx context.Context, agents []adk.Agent) (string, error) {
@@ -274,5 +271,5 @@ func deepExtrasFromConfig(ma *config.MultiAgentConfig) (outputKey string, retry
 			return prefix + "\n可用子代理（按名称 transfer / task 调用）：" + strings.Join(names, "、"), nil
 		}
 	}
-	return outputKey, retry, taskDesc
+	return outputKey, taskDesc
 }

internal/multiagent/eino_orchestration.go

@@ -94,24 +94,20 @@ func NewPlanExecuteRoot(ctx context.Context, a *PlanExecuteRootArgs) (adk.Resuma
 	if a.SkillMiddleware != nil {
 		execHandlers = append(execHandlers, a.SkillMiddleware)
 	}
-	// 4. summarization（最后，与 Deep/Supervisor 一致）
+	// 4. pre-summarization normalize + continuation dedup, then summarization (与 Deep/Supervisor 一致)
 	if a.AppCfg != nil {
 		sumMw, sumErr := newEinoSummarizationMiddleware(ctx, a.ExecModel, a.AppCfg, a.MwCfg, a.ConversationID, a.DB, a.ProjectID, a.Logger)
 		if sumErr != nil {
 			return nil, fmt.Errorf("plan_execute executor summarization: %w", sumErr)
 		}
-		execHandlers = append(execHandlers, sumMw)
-	}
-	// 5. 孤儿 tool 消息兜底：必须挂在所有改写历史中间件（summarization/reduction/skill）之后、
-	//    telemetry 之前，保证送入 ChatModel 的消息序列 tool_call ↔ tool_result 配对完整。
-	execHandlers = append(execHandlers, newOrphanToolPrunerMiddleware(a.Logger, "plan_execute_executor"))
-	if teleMw := newEinoModelInputTelemetryMiddleware(a.Logger, a.ModelName, a.ConversationID, "plan_execute_executor"); teleMw != nil {
-		execHandlers = append(execHandlers, teleMw)
-	}
-	if a.ModelFacingTrace != nil {
-		if capMw := newModelFacingTraceMiddleware(a.ModelFacingTrace); capMw != nil {
-			execHandlers = append(execHandlers, capMw)
-		}
+		execHandlers = appendEinoChatModelTailMiddlewares(execHandlers, einoChatModelTailConfig{
+			logger:         a.Logger,
+			phase:          "plan_execute_executor",
+			summarization:  sumMw,
+			modelName:      a.ModelName,
+			conversationID: a.ConversationID,
+			trace:          a.ModelFacingTrace,
+		})
 	}
 	executor, err := newPlanExecuteExecutor(ctx, &planexecute.ExecutorConfig{
 		Model:         a.ExecModel,

internal/multiagent/eino_single_runner.go

@@ -81,7 +81,7 @@ func RunEinoSingleChatModelAgent(
 	}
 
 	toolInvokeNotify := einomcp.NewToolInvokeNotifyHolder()
-	einoExecMonitor := newEinoExecuteMonitorCallback(ag, recorder)
+	einoExecBegin, einoExecFinish := newEinoExecuteMonitorCallbacks(ag, recorder)
 	mainDefs := ag.ToolsForRole(roleTools)
 	mainTools, err := einomcp.ToolsFromDefinitions(ag, holder, mainDefs, recorder, nil, toolInvokeNotify, einoSingleAgentName)
 	if err != nil {
@@ -136,7 +136,7 @@ func RunEinoSingleChatModelAgent(
 	}
 	if einoSkillMW != nil {
 		if einoFSTools && einoLoc != nil {
-			fsMw, fsErr := subAgentFilesystemMiddleware(ctx, einoLoc, toolInvokeNotify, einoSingleAgentName, einoExecMonitor, agentToolTimeoutMinutes(appCfg), nil)
+			fsMw, fsErr := subAgentFilesystemMiddleware(ctx, einoLoc, toolInvokeNotify, einoSingleAgentName, einoExecBegin, einoExecFinish, agentToolTimeoutMinutes(appCfg), agentShellNoOutputTimeoutSeconds(appCfg), nil)
 			if fsErr != nil {
 				return nil, fmt.Errorf("eino single filesystem 中间件: %w", fsErr)
 			}
@@ -144,13 +144,14 @@ func RunEinoSingleChatModelAgent(
 		}
 		handlers = append(handlers, einoSkillMW)
 	}
-	handlers = append(handlers, mainSumMw)
-	if teleMw := newEinoModelInputTelemetryMiddleware(logger, appCfg.OpenAI.Model, conversationID, "eino_single"); teleMw != nil {
-		handlers = append(handlers, teleMw)
-	}
-	if capMw := newModelFacingTraceMiddleware(modelFacingTrace); capMw != nil {
-		handlers = append(handlers, capMw)
-	}
+	handlers = appendEinoChatModelTailMiddlewares(handlers, einoChatModelTailConfig{
+		logger:         logger,
+		phase:          "eino_single",
+		summarization:  mainSumMw,
+		modelName:      appCfg.OpenAI.Model,
+		conversationID: conversationID,
+		trace:          modelFacingTrace,
+	})
 
 	maxIter := agentMaxIterations(appCfg)
 
@@ -183,18 +184,16 @@ func RunEinoSingleChatModelAgent(
 		Name:          einoSingleAgentName,
 		Description:   "Eino ADK ChatModelAgent with MCP tools for authorized security testing.",
 		Instruction:   ins,
+		GenModelInput: literalInstructionGenModelInput,
 		Model:         mainModel,
 		ToolsConfig:   mainToolsCfg,
 		MaxIterations: maxIter,
 		Handlers:      handlers,
 	}
-	outKey, modelRetry, _ := deepExtrasFromConfig(ma)
+	outKey, _ := deepExtrasFromConfig(ma)
 	if outKey != "" {
 		chatCfg.OutputKey = outKey
 	}
-	if modelRetry != nil {
-		chatCfg.ModelRetryConfig = modelRetry
-	}
 
 	chatAgent, err := adk.NewChatModelAgent(ctx, chatCfg)
 	if err != nil {

internal/multiagent/eino_skills.go

@@ -9,6 +9,7 @@ import (
 
 	"cyberstrike-ai/internal/config"
 	"cyberstrike-ai/internal/einomcp"
+	"cyberstrike-ai/internal/security"
 
 	localbk "github.com/cloudwego/eino-ext/adk/backend/local"
 	"github.com/cloudwego/eino/adk"
@@ -81,8 +82,10 @@ func subAgentFilesystemMiddleware(
 	loc *localbk.Local,
 	invokeNotify *einomcp.ToolInvokeNotifyHolder,
 	einoAgentName string,
-	recordMonitor func(toolCallID, command, stdout string, success bool, invokeErr error),
+	beginMonitor func(toolCallID, command string) string,
+	finishMonitor func(executionID, toolCallID, command, stdout string, success bool, invokeErr error),
 	toolTimeoutMinutes int,
+	shellNoOutputTimeoutSec int,
 	outputChunk func(toolName, toolCallID, chunk string),
 ) (adk.ChatModelAgentMiddleware, error) {
 	if loc == nil {
@@ -91,12 +94,14 @@ func subAgentFilesystemMiddleware(
 	return filesystem.New(ctx, &filesystem.MiddlewareConfig{
 		Backend: loc,
 		StreamingShell: &einoStreamingShellWrap{
-			inner:              loc,
-			invokeNotify:       invokeNotify,
-			einoAgentName:      strings.TrimSpace(einoAgentName),
-			outputChunk:        outputChunk,
-			recordMonitor:      recordMonitor,
-			toolTimeoutMinutes: toolTimeoutMinutes,
+			inner:                   security.NewEinoStreamingShell(),
+			invokeNotify:            invokeNotify,
+			einoAgentName:           strings.TrimSpace(einoAgentName),
+			outputChunk:             outputChunk,
+			beginMonitor:            beginMonitor,
+			finishMonitor:           finishMonitor,
+			toolTimeoutMinutes:      toolTimeoutMinutes,
+			shellNoOutputTimeoutSec: shellNoOutputTimeoutSec,
 		},
 	})
 }
@@ -108,3 +113,18 @@ func agentToolTimeoutMinutes(cfg *config.Config) int {
 	}
 	return cfg.Agent.ToolTimeoutMinutes
 }
+
+// agentShellNoOutputTimeoutSeconds：0=默认 300s（5 分钟）；-1=关闭；>0=自定义秒数。
+func agentShellNoOutputTimeoutSeconds(cfg *config.Config) int {
+	if cfg == nil {
+		return 300
+	}
+	v := cfg.Agent.ShellNoOutputTimeoutSeconds
+	if v < 0 {
+		return 0
+	}
+	if v == 0 {
+		return 300
+	}
+	return v
+}

internal/multiagent/eino_summarize.go

@@ -22,8 +22,6 @@ import (
 	"go.uber.org/zap"
 )
 
-const defaultSummarizationRetryMax = 3
-
 // einoSummarizeUserInstruction：压缩历史时保留渗透测试关键信息。
 const einoSummarizeUserInstruction = `在保持所有关键安全测试信息完整的前提下压缩对话历史。
 
@@ -97,10 +95,8 @@ func newEinoSummarizationMiddleware(
 		}
 	}
 
-	retryMax := defaultSummarizationRetryMax
-	if mwCfg != nil && mwCfg.SummarizationRetryMaxAttempts > 0 {
-		retryMax = mwCfg.SummarizationRetryMaxAttempts
-	}
+	retryPolicy := einoTransientRunRetryPolicyFromMW(mwCfg)
+	retryMax := retryPolicy.maxAttempts
 
 	// ModelOptions apply only to summarization Generate (same ChatModel instance as the agent).
 	// Strip thinking/reasoning on this call path; mark requests for empty-choices diagnostics.
@@ -137,13 +133,14 @@ func newEinoSummarizationMiddleware(
 		Retry: &summarization.RetryConfig{
 			MaxRetries: &retryMax,
 			ShouldRetry: func(_ context.Context, _ adk.Message, err error) bool {
-				if err != nil && logger != nil {
-					logger.Warn("eino summarization generate attempt failed, will retry if attempts remain",
+				retry := isEinoTransientRunError(err)
+				if retry && logger != nil {
+					logger.Warn("eino summarization generate transient error, will retry if attempts remain",
 						zap.Error(err),
 						zap.Int("max_retries", retryMax),
 					)
 				}
-				return err != nil
+				return retry
 			},
 		},
 		Finalize: func(ctx context.Context, originalMessages []adk.Message, summary adk.Message) ([]adk.Message, error) {
@@ -153,6 +150,7 @@ func newEinoSummarizationMiddleware(
 			}
 			if appCfg != nil {
 				out = refreshFactIndexInMessages(out, db, projectID, appCfg.Project, logger)
+				out = refreshUserVerbatimAnchorInMessages(out, db, conversationID, appCfg.MultiAgent.UserVerbatimAnchorMaxRunesEffective(), logger)
 			}
 			return out, nil
 		},
@@ -260,17 +258,19 @@ func summarizeFinalizeWithRecentAssistantToolTrail(
 		nonSystem = append(nonSystem, msg)
 	}
 
+	mergedSystem := mergeCollectedSystemMessages(systemMsgs)
+
 	if recentTrailTokenBudget <= 0 || len(nonSystem) == 0 {
-		out := make([]adk.Message, 0, len(systemMsgs)+1)
-		out = append(out, systemMsgs...)
+		out := make([]adk.Message, 0, len(mergedSystem)+1)
+		out = append(out, mergedSystem...)
 		out = append(out, summary)
 		return out, nil
 	}
 
 	rounds := splitMessagesIntoRounds(nonSystem)
 	if len(rounds) == 0 {
-		out := make([]adk.Message, 0, len(systemMsgs)+1)
-		out = append(out, systemMsgs...)
+		out := make([]adk.Message, 0, len(mergedSystem)+1)
+		out = append(out, mergedSystem...)
 		out = append(out, summary)
 		return out, nil
 	}
@@ -322,8 +322,8 @@ func summarizeFinalizeWithRecentAssistantToolTrail(
 		selectedMsgs = append(selectedMsgs, selectedRoundsReverse[i].messages...)
 	}
 
-	out := make([]adk.Message, 0, len(systemMsgs)+1+len(selectedMsgs))
-	out = append(out, systemMsgs...)
+	out := make([]adk.Message, 0, len(mergedSystem)+1+len(selectedMsgs))
+	out = append(out, mergedSystem...)
 	out = append(out, summary)
 	out = append(out, selectedMsgs...)
 	return out, nil
@@ -414,6 +414,36 @@ func writeSummarizationTranscript(path string, msgs []adk.Message) error {
 	return nil
 }
 
+// refreshUserVerbatimAnchorInMessages 压缩后从 messages 表刷新 system 中的用户原文锚点。
+func refreshUserVerbatimAnchorInMessages(msgs []adk.Message, db *database.DB, conversationID string, maxRunes int, logger *zap.Logger) []adk.Message {
+	if maxRunes < 0 || db == nil {
+		return msgs
+	}
+	conversationID = strings.TrimSpace(conversationID)
+	if conversationID == "" {
+		return msgs
+	}
+	rows, err := db.GetMessages(conversationID)
+	if err != nil {
+		if logger != nil {
+			logger.Warn("summarization: 刷新用户原文锚点失败",
+				zap.String("conversationId", conversationID),
+				zap.Error(err),
+			)
+		}
+		return msgs
+	}
+	block := project.BuildUserVerbatimAnchorBlockFromMessages(rows, maxRunes)
+	if block == "" {
+		return msgs
+	}
+	out := project.RefreshUserVerbatimAnchorInMessages(msgs, block)
+	if logger != nil {
+		logger.Info("summarization: 已刷新用户原文锚点", zap.String("conversationId", conversationID))
+	}
+	return out
+}
+
 func einoSummarizationTokenCounter(openAIModel string) summarization.TokenCounterFunc {
 	tc := agent.NewTikTokenCounter()
 	return func(ctx context.Context, input *summarization.TokenCounterInput) (int, error) {

internal/multiagent/eino_summarize_payload.go

@@ -1,35 +1,12 @@
 package multiagent
 
 import (
-	"github.com/bytedance/sonic"
+	copenai "cyberstrike-ai/internal/openai"
 )
 
 // stripReasoningFromSummarizationPayload removes thinking / reasoning fields from a
 // chat-completions JSON body. Applied only to summarization Generate calls via
 // model.ModelOptions on the shared ChatModel — main-agent requests are unchanged.
 func stripReasoningFromSummarizationPayload(rawBody []byte) ([]byte, error) {
-	var payload map[string]any
-	if err := sonic.Unmarshal(rawBody, &payload); err != nil {
-		return rawBody, nil
-	}
-	changed := false
-	for _, key := range []string{
-		"thinking",
-		"reasoning_effort",
-		"output_config",
-		"reasoning",
-	} {
-		if _, ok := payload[key]; ok {
-			delete(payload, key)
-			changed = true
-		}
-	}
-	if !changed {
-		return rawBody, nil
-	}
-	out, err := sonic.Marshal(payload)
-	if err != nil {
-		return rawBody, err
-	}
-	return out, nil
+	return copenai.StripReasoningFromChatCompletionBody(rawBody)
 }

internal/multiagent/eino_summarize_test.go

@@ -192,8 +192,8 @@ func TestSummarizeFinalize_KeepsToolRoundIntact(t *testing.T) {
 	if len(out) < 2 {
 		t.Fatalf("output too short: %d", len(out))
 	}
-	if out[0] != sys {
-		t.Fatalf("first message must be system")
+	if out[0].Role != schema.System || out[0].Content != "sys" {
+		t.Fatalf("first message must be system sys, got %s: %q", out[0].Role, out[0].Content)
 	}
 	if out[1] != summary {
 		t.Fatalf("second message must be summary")
@@ -293,12 +293,12 @@ func TestSummarizeFinalize_BudgetZeroFallsBackToSummaryOnly(t *testing.T) {
 	if err != nil {
 		t.Fatalf("unexpected error: %v", err)
 	}
-	if len(out) != 2 || out[0] != sys || out[1] != summary {
+	if len(out) != 2 || out[0].Role != schema.System || out[0].Content != "sys" || out[1] != summary {
 		t.Fatalf("budget=0 must yield [system, summary] only, got %+v", out)
 	}
 }
 
-func TestSummarizeFinalize_PreservesAllSystemMessages(t *testing.T) {
+func TestSummarizeFinalize_MergesSystemMessages(t *testing.T) {
 	sys1 := schema.SystemMessage("sys1")
 	sys2 := schema.SystemMessage("sys2")
 	summary := schema.AssistantMessage("s", nil)
@@ -321,10 +321,13 @@ func TestSummarizeFinalize_PreservesAllSystemMessages(t *testing.T) {
 	for _, m := range out {
 		if m != nil && m.Role == schema.System {
 			systemCount++
+			if got := m.Content; got != "sys1\n\nsys2" {
+				t.Fatalf("unexpected merged system content: %q", got)
+			}
 		}
 	}
-	if systemCount != 2 {
-		t.Fatalf("want 2 system messages retained, got %d", systemCount)
+	if systemCount != 1 {
+		t.Fatalf("want 1 merged system message, got %d", systemCount)
 	}
 }
 
@@ -378,6 +381,12 @@ func TestWriteSummarizationTranscript(t *testing.T) {
 	if !strings.Contains(text, "tool_calls:") || !strings.Contains(text, "nmap output") {
 		t.Fatalf("missing tool round: %q", text)
 	}
+	if !strings.Contains(text, `"name":"stub_tool"`) || !strings.Contains(text, `"arguments":"{}"`) {
+		t.Fatalf("missing tool name/arguments: %q", text)
+	}
+	if strings.Contains(text, "tool_call_id") || strings.Contains(text, `"id":"tc1"`) {
+		t.Fatalf("transcript should omit tool_call_id: %q", text)
+	}
 }
 
 func TestSanitizeSystemContentForTranscript_BestPractice(t *testing.T) {
@@ -400,9 +409,9 @@ func TestSanitizeSystemContentForTranscript_BestPractice(t *testing.T) {
 		"需要写入请使用 upsert_project_fact。",
 		project.FactIndexSectionEndMarker,
 		"",
-		"# Skills System",
-		"**How to Use Skills**",
-		"Remember: Skills make you more capable",
+		transcriptSkillsSystemMarker,
+		"**如何使用 Skill（技能）（渐进式展示）：**",
+		"记住：Skill 让你更加强大和稳定",
 	}, "\n")
 
 	out := sanitizeSystemContentForTranscript(system)
@@ -412,7 +421,7 @@ func TestSanitizeSystemContentForTranscript_BestPractice(t *testing.T) {
 	if strings.Contains(out, "- nmap") || strings.Contains(out, "高强度扫描要求") {
 		t.Fatalf("static persona should be stripped: %q", out)
 	}
-	if strings.Contains(out, "# Skills System") || strings.Contains(out, "How to Use Skills") {
+	if strings.Contains(out, transcriptSkillsSystemMarker) || strings.Contains(out, "如何使用 Skill") {
 		t.Fatalf("skills boilerplate should be stripped: %q", out)
 	}
 	if !strings.Contains(out, transcriptStaticSystemOmitNote) {
@@ -426,7 +435,7 @@ func TestSanitizeSystemContentForTranscript_BestPractice(t *testing.T) {
 func TestFormatSummarizationTranscript_OmitsBloatedSystem(t *testing.T) {
 	t.Parallel()
 	msgs := []adk.Message{
-		schema.SystemMessage("以下是当前会话绑定的工具名称索引\n- nmap\n\n你是CyberStrikeAI\n" + project.FactIndexSectionStartMarker + "\n## 项目黑板索引（project: p1, id: x）\n（暂无事实）\n" + project.FactIndexSectionEndMarker + "\n# Skills System\nboiler"),
+		schema.SystemMessage("以下是当前会话绑定的工具名称索引\n- nmap\n\n你是CyberStrikeAI\n" + project.FactIndexSectionStartMarker + "\n## 项目黑板索引（project: p1, id: x）\n（暂无事实）\n" + project.FactIndexSectionEndMarker + "\n" + transcriptSkillsSystemMarker + "\nboiler"),
 		schema.UserMessage("hello"),
 		schema.AssistantMessage("reply", nil),
 	}

internal/multiagent/eino_summarize_transcript.go

@@ -20,9 +20,16 @@ const (
 	transcriptStaticSystemOmitNote = "[static system prompt omitted — unchanged in live context after compaction]"
 	transcriptToolIndexStartMarker   = "以下是当前会话绑定的工具名称索引"
 	transcriptPersonaStartMarker     = "你是CyberStrikeAI"
-	transcriptSkillsSystemMarker     = "# Skills System"
+	// ADK LanguageChinese injects skill middleware prompt with this header (see eino adk/middlewares/skill/prompt.go).
+	transcriptSkillsSystemMarker        = "# Skill 系统"
+	transcriptSkillsSystemMarkerEnglish = "# Skills System"
 )
 
+type transcriptToolCall struct {
+	Name      string `json:"name"`
+	Arguments string `json:"arguments"`
+}
+
 // formatSummarizationTranscript renders pre-compaction messages for transcript.txt.
 // Best practice: keep full user/assistant/tool turns; slim system to dynamic blocks only.
 func formatSummarizationTranscript(msgs []adk.Message) string {
@@ -81,13 +88,23 @@ func stripToolNamesIndexFromSystem(s string) string {
 }
 
 func stripSkillsSystemBoilerplate(s string) string {
-	idx := strings.Index(s, transcriptSkillsSystemMarker)
+	idx := indexFirstSubstring(s, transcriptSkillsSystemMarker, transcriptSkillsSystemMarkerEnglish)
 	if idx < 0 {
 		return strings.TrimSpace(s)
 	}
 	return strings.TrimSpace(s[:idx])
 }
 
+func indexFirstSubstring(s string, markers ...string) int {
+	first := -1
+	for _, m := range markers {
+		if i := strings.Index(s, m); i >= 0 && (first < 0 || i < first) {
+			first = i
+		}
+	}
+	return first
+}
+
 func extractProjectBlackboardSection(s string) string {
 	start := strings.Index(s, project.FactIndexSectionStartMarker)
 	if start < 0 {
@@ -138,15 +155,21 @@ func appendTranscriptMessage(sb *strings.Builder, msg adk.Message) {
 		}
 	}
 	if len(msg.ToolCalls) > 0 {
-		if b, err := sonic.Marshal(msg.ToolCalls); err == nil {
+		if b, err := sonic.Marshal(formatTranscriptToolCalls(msg.ToolCalls)); err == nil {
 			sb.WriteString("tool_calls: ")
 			sb.Write(b)
 			sb.WriteByte('\n')
 		}
 	}
-	if msg.ToolCallID != "" {
-		sb.WriteString("tool_call_id: ")
-		sb.WriteString(msg.ToolCallID)
-		sb.WriteByte('\n')
-	}
+}
+
+func formatTranscriptToolCalls(calls []schema.ToolCall) []transcriptToolCall {
+	out := make([]transcriptToolCall, 0, len(calls))
+	for _, tc := range calls {
+		out = append(out, transcriptToolCall{
+			Name:      tc.Function.Name,
+			Arguments: tc.Function.Arguments,
+		})
+	}
+	return out
 }

internal/multiagent/eino_tool_name_injection.go

@@ -46,6 +46,10 @@ func injectToolNamesOnlyInstruction(ctx context.Context, instruction string, too
 		sb.WriteString("2) 调用具体工具前，请先确认该工具的参数要求（以当前请求中的工具定义为准）；不确定时先澄清再调用。\n")
 		sb.WriteString("3) 不要臆造不存在的工具名。\n\n")
 	}
+	if s := strings.TrimSpace(injectShellToolGuidance("", names)); s != "" {
+		sb.WriteString(s)
+		sb.WriteString("\n\n")
+	}
 	if s := strings.TrimSpace(instruction); s != "" {
 		sb.WriteString(s)
 	}

internal/multiagent/eino_transient_retry.go

@@ -3,6 +3,7 @@ package multiagent
 import (
 	"context"
 	"errors"
+	"fmt"
 	"strings"
 	"time"
 
@@ -17,8 +18,9 @@ const (
 	defaultEinoRunRetryMaxBackoff  = 30 * time.Second
 )
 
-// isEinoTransientRunError 判断 ADK 运行期错误是否适合指数退避续跑（429、5xx、网络抖动等）。
-// 用户取消、超时、迭代上限等由 run loop 单独处理，不在此列。
+// isEinoTransientRunError 是 Eino 运行期「可退避重试 vs 直接失败」的唯一判据。
+// 429/5xx/网络抖动等返回 true；用户取消、超时、迭代上限、鉴权失败等返回 false。
+// 其它模块（run loop、summarization 等）只调用本函数，不在别处维护平行规则。
 func isEinoTransientRunError(err error) bool {
 	if err == nil {
 		return false
@@ -60,6 +62,7 @@ func isEinoTransientRunError(err error) bool {
 		"dial tcp",
 		"tls handshake timeout",
 		"stream error",
+		"goaway", // http2: server sent GOAWAY and closed the connection
 		"unexpected eof",
 		`": eof`, // net/http: Post "url": EOF (often wraps io.EOF)
 		"unexpected end of json",
@@ -78,6 +81,71 @@ func isEinoTransientRunError(err error) bool {
 	return false
 }
 
+type einoTransientRunRetryPolicy struct {
+	maxAttempts int
+	maxBackoff  time.Duration
+}
+
+func einoTransientRunRetryPolicyFromArgs(args *einoADKRunLoopArgs) einoTransientRunRetryPolicy {
+	return einoTransientRunRetryPolicy{
+		maxAttempts: einoRunRetryMaxAttempts(args),
+		maxBackoff:  einoRunRetryMaxBackoff(args),
+	}
+}
+
+func einoTransientRunRetryPolicyFromMW(mw *config.MultiAgentEinoMiddlewareConfig) einoTransientRunRetryPolicy {
+	maxBackoff := defaultEinoRunRetryMaxBackoff
+	if mw != nil && mw.RunRetryMaxBackoffSec > 0 {
+		maxBackoff = time.Duration(mw.RunRetryMaxBackoffSec) * time.Second
+	}
+	return einoTransientRunRetryPolicy{
+		maxAttempts: RunRetryMaxAttemptsFromConfig(mw),
+		maxBackoff:  maxBackoff,
+	}
+}
+
+// einoTransientRunRetrier 在 run loop 内对临时错误做指数退避并重启 Runner（唯一重试执行层）。
+type einoTransientRunRetrier struct {
+	policy   einoTransientRunRetryPolicy
+	attempts int
+}
+
+func newEinoTransientRunRetrier(policy einoTransientRunRetryPolicy) *einoTransientRunRetrier {
+	return &einoTransientRunRetrier{policy: policy}
+}
+
+// tryRetry 对临时错误退避后返回重启消息；次数用尽返回 exhausted 错误。
+func (r *einoTransientRunRetrier) tryRetry(
+	ctx context.Context,
+	runErr error,
+	args *einoADKRunLoopArgs,
+	baseMsgs, accumulated []adk.Message,
+	baseCount int,
+) (restarted bool, restartMsgs []adk.Message, ctxSource einoRunRestartContextSource, backoff time.Duration, fatal error) {
+	if runErr == nil || !isEinoTransientRunError(runErr) {
+		return false, nil, "", 0, runErr
+	}
+	r.attempts++
+	if r.attempts > r.policy.maxAttempts {
+		return false, nil, "", 0, fmt.Errorf("transient retry exhausted after %d attempts: %w", r.policy.maxAttempts, runErr)
+	}
+	backoff = einoTransientRetryBackoff(r.attempts-1, r.policy.maxBackoff)
+	select {
+	case <-ctx.Done():
+		return false, nil, "", 0, ctx.Err()
+	case <-time.After(backoff):
+	}
+	restartMsgs, ctxSource = einoMessagesForRunRestart(args, baseMsgs, accumulated, baseCount)
+	return true, restartMsgs, ctxSource, backoff, nil
+}
+
+func (r *einoTransientRunRetrier) attempt() int { return r.attempts }
+
+func (r *einoTransientRunRetrier) maxAttempts() int { return r.policy.maxAttempts }
+
+// reset 在退避重试后成功推进（流/消息完整接收）时清零计数，使后续临时错误从第 1 次退避重新开始。
+func (r *einoTransientRunRetrier) reset() { r.attempts = 0 }
+
 func einoRunRetryMaxAttempts(args *einoADKRunLoopArgs) int {
 	if args != nil && args.RunRetryMaxAttempts > 0 {
 		return args.RunRetryMaxAttempts
@@ -85,7 +153,7 @@ func einoRunRetryMaxAttempts(args *einoADKRunLoopArgs) int {
 	return defaultEinoRunRetryMaxAttempts
 }
 
-// RunRetryMaxAttemptsFromConfig 供 handler 分段续跑计数（与 eino_middleware.run_retry_max_attempts 一致）。
+// RunRetryMaxAttemptsFromConfig 与 eino_middleware.run_retry_max_attempts 一致。
 func RunRetryMaxAttemptsFromConfig(mw *config.MultiAgentEinoMiddlewareConfig) int {
 	if mw != nil && mw.RunRetryMaxAttempts > 0 {
 		return mw.RunRetryMaxAttempts
@@ -93,15 +161,6 @@ func RunRetryMaxAttemptsFromConfig(mw *config.MultiAgentEinoMiddlewareConfig) in
 	return defaultEinoRunRetryMaxAttempts
 }
 
-// TransientRetryBackoff 供 handler 在分段续跑前退避。
-func TransientRetryBackoff(attempt int, maxBackoffSec int) time.Duration {
-	max := defaultEinoRunRetryMaxBackoff
-	if maxBackoffSec > 0 {
-		max = time.Duration(maxBackoffSec) * time.Second
-	}
-	return einoTransientRetryBackoff(attempt, max)
-}
-
 func einoRunRetryMaxBackoff(args *einoADKRunLoopArgs) time.Duration {
 	if args != nil && args.RunRetryMaxBackoffSec > 0 {
 		return time.Duration(args.RunRetryMaxBackoffSec) * time.Second
@@ -122,37 +181,35 @@ const (
 // 1) ModelFacingTrace（与模型实际入参一致） 2) 事件流累积的 runAccumulatedMsgs 3) 初始 msgs。
 func einoMessagesForRunRestart(args *einoADKRunLoopArgs, baseMsgs, accumulated []adk.Message, baseCount int) ([]adk.Message, einoRunRestartContextSource) {
 	if trace := persistTraceSource(args, nil); len(trace) > 0 {
-		return append([]adk.Message(nil), trace...), einoRestartContextModelTrace
+		// modelFacingTrace includes prior Instruction system message(s); genModelInput will prepend again.
+		return stripADKSystemMessages(trace), einoRestartContextModelTrace
 	}
 	if len(accumulated) > baseCount {
-		return append([]adk.Message(nil), accumulated...), einoRestartContextAccumulated
+		return stripADKSystemMessages(accumulated), einoRestartContextAccumulated
 	}
 	return append([]adk.Message(nil), baseMsgs...), einoRestartContextInitial
 }
 
-// adkMessagesHasUserContent 从尾部向前查找，是否已有与 want 相同的 user 消息（避免重复 append）。
+// adkMessagesHasUserContent reports whether the conversation tail is already a user turn
+// with the given content. Only the last message counts: matching text in an earlier round
+// (e.g. user repeats the same prompt after an assistant reply) must not suppress appending
+// the new user turn — Claude 4.6+ rejects requests whose final message is assistant.
 func adkMessagesHasUserContent(msgs []adk.Message, want string) bool {
 	want = strings.TrimSpace(want)
 	if want == "" {
 		return true
 	}
-	for i := len(msgs) - 1; i >= 0; i-- {
-		m := msgs[i]
-		if m == nil {
-			continue
-		}
-		if m.Role == schema.User {
-			return strings.TrimSpace(m.Content) == want
-		}
-		if m.Role == schema.Assistant || m.Role == schema.Tool {
-			continue
-		}
-		break
+	if len(msgs) == 0 {
+		return false
 	}
-	return false
+	last := msgs[len(msgs)-1]
+	if last == nil || last.Role != schema.User {
+		return false
+	}
+	return strings.TrimSpace(last.Content) == want
 }
 
-// appendUserMessageIfNeeded 在 history 轨迹之后追加本轮 user 消息（仅当轨迹中尚未包含该句）。
+// appendUserMessageIfNeeded 在 history 轨迹之后追加本轮 user 消息（仅当尾部已是相同 user 句）。
 func appendUserMessageIfNeeded(msgs []adk.Message, userMessage string) []adk.Message {
 	if strings.TrimSpace(userMessage) == "" || adkMessagesHasUserContent(msgs, userMessage) {
 		return msgs

internal/multiagent/eino_transient_retry_test.go

@@ -27,6 +27,7 @@ func TestIsEinoTransientRunError(t *testing.T) {
 		{"429", errors.New("HTTP 429 Too Many Requests"), true},
 		{"rate limit", errors.New(`{"error":"rate limit exceeded"}`), true},
 		{"connection reset", errors.New("read tcp: connection reset by peer"), true},
+		{"http2 goaway", errors.New("failed to receive stream chunk: error, http2: server sent GOAWAY and closed the connection; LastStreamID=791, ErrCode=NO_ERROR"), true},
 		{"unexpected eof", errors.New("unexpected EOF"), true},
 		{"503", errors.New("upstream returned 503"), true},
 		{"iteration limit", errors.New("max iteration reached"), false},
@@ -90,6 +91,46 @@ func TestEinoRunRetryMaxAttemptsFromArgs(t *testing.T) {
 	}
 }
 
+func TestEinoTransientRunRetrierReset(t *testing.T) {
+	t.Parallel()
+	r := newEinoTransientRunRetrier(einoTransientRunRetryPolicy{maxAttempts: 10, maxBackoff: 30 * time.Second})
+	r.attempts = 3
+	r.reset()
+	if r.attempt() != 0 {
+		t.Fatalf("after reset: attempt=%d, want 0", r.attempt())
+	}
+	// 重置后下一次退避应从 2s 起算（attempt index 0）。
+	if got := einoTransientRetryBackoff(r.attempt(), r.policy.maxBackoff); got != 2*time.Second {
+		t.Fatalf("backoff after reset: got %v, want 2s", got)
+	}
+}
+
+func TestEinoTransientRunRetrierConsecutiveFailures(t *testing.T) {
+	t.Parallel()
+	r := newEinoTransientRunRetrier(einoTransientRunRetryPolicy{maxAttempts: 10, maxBackoff: 30 * time.Second})
+	ctx := context.Background()
+	runErr := errors.New("internal server error")
+	args := &einoADKRunLoopArgs{}
+	base := []adk.Message{schema.UserMessage("hi")}
+
+	for want := 1; want <= 3; want++ {
+		restarted, _, _, _, err := r.tryRetry(ctx, runErr, args, base, nil, len(base))
+		if err != nil {
+			t.Fatalf("tryRetry attempt %d: %v", want, err)
+		}
+		if !restarted {
+			t.Fatalf("tryRetry attempt %d: want restarted", want)
+		}
+		if got := r.attempt(); got != want {
+			t.Fatalf("after failure %d: attempt=%d, want %d", want, got, want)
+		}
+	}
+	r.reset()
+	if r.attempt() != 0 {
+		t.Fatalf("after successful recovery reset: attempt=%d, want 0", r.attempt())
+	}
+}
+
 func TestAppendUserMessageIfNeeded(t *testing.T) {
 	t.Parallel()
 	msgs := []adk.Message{schema.UserMessage("old task")}
@@ -103,9 +144,17 @@ func TestAppendUserMessageIfNeeded(t *testing.T) {
 	}
 }
 
-func TestErrTransientRetryContinue(t *testing.T) {
+func TestAppendUserMessageIfNeeded_repeatPromptAfterAssistant(t *testing.T) {
 	t.Parallel()
-	if !errors.Is(ErrTransientRetryContinue, ErrTransientRetryContinue) {
-		t.Fatal("sentinel should match")
+	msgs := []adk.Message{
+		schema.UserMessage("扫描 example.com"),
+		schema.AssistantMessage("开始扫描...", nil),
+	}
+	out := appendUserMessageIfNeeded(msgs, "扫描 example.com")
+	if len(out) != 3 {
+		t.Fatalf("should append new user turn after assistant reply: len=%d", len(out))
+	}
+	if out[2].Role != schema.User || out[2].Content != "扫描 example.com" {
+		t.Fatalf("tail should be repeated user prompt, got role=%s content=%q", out[2].Role, out[2].Content)
 	}
 }

internal/multiagent/execute_exit_error.go

@@ -0,0 +1,15 @@
+package multiagent
+
+import "fmt"
+
+// ExecuteExitError 表示 execute 命令非零退出（预期失败，非超时/中断/流异常）。
+type ExecuteExitError struct {
+	Code int
+}
+
+func (e *ExecuteExitError) Error() string {
+	if e == nil {
+		return "exit status unknown"
+	}
+	return fmt.Sprintf("exit status %d", e.Code)
+}

internal/multiagent/gen_model_input.go

@@ -0,0 +1,23 @@
+package multiagent
+
+import (
+	"context"
+
+	"github.com/cloudwego/eino/adk"
+	"github.com/cloudwego/eino/schema"
+)
+
+// literalInstructionGenModelInput passes Instruction through as a system message without
+// FString template formatting. Eino defaultGenModelInput formats instruction whenever
+// SessionValues exist; prompts with literal curly braces (project blackboard "{关系边: ...}",
+// JSON examples, link syntax) then fail with "could not find key".
+//
+// Matches eino/adk/prebuilt/deep genModelInput — the supported fix per Eino docs.
+func literalInstructionGenModelInput(ctx context.Context, instruction string, input *adk.AgentInput) ([]adk.Message, error) {
+	msgs := make([]adk.Message, 0, len(input.Messages)+1)
+	if instruction != "" {
+		msgs = append(msgs, schema.SystemMessage(instruction))
+	}
+	msgs = append(msgs, input.Messages...)
+	return msgs, nil
+}

internal/multiagent/gen_model_input_test.go

@@ -0,0 +1,33 @@
+package multiagent
+
+import (
+	"context"
+	"strings"
+	"testing"
+
+	"github.com/cloudwego/eino/adk"
+	"github.com/cloudwego/eino/schema"
+)
+
+func TestLiteralInstructionGenModelInput_PreservesLiteralCurlyBraces(t *testing.T) {
+	t.Parallel()
+	instruction := "- [finding/x] summary {关系边: discovered_on←target/dev}\n" +
+		"如 finding 上 {from:target/*, type:discovered_on}"
+	msgs, err := literalInstructionGenModelInput(context.Background(), instruction, &adk.AgentInput{
+		Messages: []adk.Message{schema.UserMessage("继续")},
+	})
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if len(msgs) != 2 {
+		t.Fatalf("expected 2 messages, got %d", len(msgs))
+	}
+	if msgs[0].Role != schema.System {
+		t.Fatalf("first message must be system, got %s", msgs[0].Role)
+	}
+	for _, want := range []string{"{关系边:", "{from:target/*, type:discovered_on}"} {
+		if !strings.Contains(msgs[0].Content, want) {
+			t.Fatalf("system content missing %q: %q", want, msgs[0].Content)
+		}
+	}
+}

internal/multiagent/hitl_middleware.go

@@ -3,7 +3,6 @@ package multiagent
 import (
 	"context"
 	"errors"
-	"fmt"
 	"strings"
 
 	"github.com/cloudwego/eino/adk"
@@ -75,8 +74,8 @@ func hitlInvokableToolCallMiddleware() compose.InvokableToolMiddleware {
 					if err != nil {
 						if IsHumanRejectError(err) {
 							// Human rejection should be a soft tool result so the model can continue iterating.
-							msg := fmt.Sprintf("[HITL Reject] Tool '%s' was rejected by human reviewer. Reason: %s\nPlease adjust parameters/plan and continue without this call.",
-								input.Name, strings.TrimSpace(err.Error()))
+							// tool_search 须保持 JSON，否则 Eino toolsearch 中间件解析历史时会硬崩 ChatModel。
+							msg := HitlRejectToolResult(input.Name, err.Error())
 							// transfer_to_agent 在 Eino 中标记为 returnDirectly：工具成功后 ReAct 子图会直接 END，
 							// 并依赖真实工具内的 SendToolGenAction 触发移交。HITL 拒绝时不会执行真实工具，
 							// 若仍走 returnDirectly 分支，监督者会在无 Transfer 动作的情况下结束，模型不再迭代。
@@ -103,8 +102,7 @@ func hitlStreamableToolCallMiddleware() compose.StreamableToolMiddleware {
 					edited, err := fn(ctx, input.Name, input.Arguments)
 					if err != nil {
 						if IsHumanRejectError(err) {
-							msg := fmt.Sprintf("[HITL Reject] Tool '%s' was rejected by human reviewer. Reason: %s\nPlease adjust parameters/plan and continue without this call.",
-								input.Name, strings.TrimSpace(err.Error()))
+							msg := HitlRejectToolResult(input.Name, err.Error())
 							hitlClearReturnDirectlyIfTransfer(ctx, input.Name)
 							return &compose.StreamToolOutput{
 								Result: schema.StreamReaderFromArray([]string{msg}),

internal/multiagent/hitl_toolsearch_compat.go

@@ -0,0 +1,85 @@
+package multiagent
+
+import (
+	"encoding/json"
+	"fmt"
+	"strings"
+)
+
+const toolSearchToolName = "tool_search"
+
+// HitlExemptMetaTools 为编排/元工具：不直接执行攻击动作，但会阻塞 agent 控制流。
+// tool_search 必须免审批，否则其 HITL 拒绝结果与 Eino toolsearch 中间件不兼容（会硬崩 ChatModel）。
+var HitlExemptMetaTools = []string{
+	toolSearchToolName,
+	"skill",
+	"task",
+	"write_todos",
+	"transfer_to_agent",
+	"exit",
+	"TaskCreate",
+	"TaskGet",
+	"TaskUpdate",
+	"TaskList",
+}
+
+// IsToolSearchTool reports whether name is the Eino dynamictool tool_search meta-tool.
+func IsToolSearchTool(name string) bool {
+	return strings.EqualFold(strings.TrimSpace(name), toolSearchToolName)
+}
+
+// MergeHitlExemptMetaTools unions configured whitelist with built-in meta-tool exemptions.
+func MergeHitlExemptMetaTools(configured []string) []string {
+	merged := make([]string, 0, len(configured)+len(HitlExemptMetaTools))
+	seen := make(map[string]struct{}, len(configured)+len(HitlExemptMetaTools))
+	add := func(name string) {
+		n := strings.ToLower(strings.TrimSpace(name))
+		if n == "" {
+			return
+		}
+		if _, ok := seen[n]; ok {
+			return
+		}
+		seen[n] = struct{}{}
+		merged = append(merged, strings.TrimSpace(name))
+	}
+	for _, t := range configured {
+		add(t)
+	}
+	for _, t := range HitlExemptMetaTools {
+		add(t)
+	}
+	return merged
+}
+
+type toolSearchHitlRejectPayload struct {
+	SelectedTools []string `json:"selectedTools"`
+	HitlRejected  bool     `json:"_hitlRejected"`
+	Reason        string   `json:"reason"`
+}
+
+// HitlRejectToolResult returns a tool result body safe for downstream consumers.
+// tool_search must stay JSON-shaped so toolsearch.extractSelectedTools does not terminate the graph.
+func HitlRejectToolResult(toolName, reason string) string {
+	reason = strings.TrimSpace(reason)
+	if !IsToolSearchTool(toolName) {
+		if reason == "" {
+			reason = "rejected by reviewer"
+		}
+		return fmt.Sprintf("[HITL Reject] Tool '%s' was rejected by reviewer. Reason: %s\nPlease adjust parameters/plan and continue without this call.",
+			strings.TrimSpace(toolName), reason)
+	}
+	payload := toolSearchHitlRejectPayload{
+		SelectedTools: []string{},
+		HitlRejected:  true,
+		Reason:        reason,
+	}
+	if payload.Reason == "" {
+		payload.Reason = "tool_search rejected by reviewer; no dynamic tools unlocked"
+	}
+	out, err := json.Marshal(payload)
+	if err != nil {
+		return `{"selectedTools":[],"_hitlRejected":true,"reason":"tool_search rejected by reviewer"}`
+	}
+	return string(out)
+}

internal/multiagent/hitl_toolsearch_compat_test.go

@@ -0,0 +1,48 @@
+package multiagent
+
+import (
+	"encoding/json"
+	"strings"
+	"testing"
+)
+
+func TestHitlRejectToolResult_toolSearchIsJSON(t *testing.T) {
+	raw := HitlRejectToolResult("tool_search", "rejected by user: timeout")
+	var payload toolSearchHitlRejectPayload
+	if err := json.Unmarshal([]byte(raw), &payload); err != nil {
+		t.Fatalf("unmarshal: %v", err)
+	}
+	if len(payload.SelectedTools) != 0 {
+		t.Fatalf("expected empty selectedTools, got %v", payload.SelectedTools)
+	}
+	if !payload.HitlRejected {
+		t.Fatal("expected _hitlRejected true")
+	}
+	if !strings.Contains(payload.Reason, "timeout") {
+		t.Fatalf("reason=%q", payload.Reason)
+	}
+}
+
+func TestHitlRejectToolResult_otherToolKeepsLegacyText(t *testing.T) {
+	raw := HitlRejectToolResult("nmap", "too risky")
+	if strings.HasPrefix(raw, "{") {
+		t.Fatalf("expected legacy text, got %q", raw)
+	}
+	if !strings.HasPrefix(raw, "[HITL Reject]") {
+		t.Fatalf("expected [HITL Reject] prefix, got %q", raw)
+	}
+}
+
+func TestMergeHitlExemptMetaTools_includesToolSearch(t *testing.T) {
+	merged := MergeHitlExemptMetaTools([]string{"read_file"})
+	found := false
+	for _, name := range merged {
+		if IsToolSearchTool(name) {
+			found = true
+			break
+		}
+	}
+	if !found {
+		t.Fatalf("tool_search missing from %v", merged)
+	}
+}

internal/multiagent/interrupt.go

@@ -5,11 +5,3 @@ import "errors"
 // ErrInterruptContinue 作为 context.CancelCause 使用：用户选择「中断并继续」且当前无进行中的 MCP 工具时，
 // 取消当前推理/流式输出，并在同一会话任务内携带用户补充说明自动续跑下一轮（类似 Hermes 式人机回合）。
 var ErrInterruptContinue = errors.New("agent interrupt: continue with user-supplied context")
-
-// ErrTransientRetryContinue 表示 Run 因 429/网络等临时错误结束，应由 handler 落库轨迹后
-// loadHistoryFromAgentTrace 再开下一轮 Run（与 ErrInterruptContinue 同级的「分段续跑」语义）。
-var ErrTransientRetryContinue = errors.New("agent transient: retry after persisting trace")
-
-// ErrEmptyResponseContinue 表示 Eino ADK 会话正常结束但未捕获到助手正文，应由 handler 落库轨迹后
-// loadHistoryFromAgentTrace 再开下一轮 Run（与 ErrInterruptContinue / ErrTransientRetryContinue 同级）。
-var ErrEmptyResponseContinue = errors.New("agent empty response: continue after persisting trace")

internal/multiagent/orchestrator_instruction.go

@@ -6,6 +6,7 @@ import (
 	"cyberstrike-ai/internal/agents"
 	"cyberstrike-ai/internal/config"
 	"cyberstrike-ai/internal/project"
+	"cyberstrike-ai/internal/projectprompt"
 )
 
 // DefaultPlanExecuteOrchestratorInstruction 当未配置 plan_execute 专用 Markdown / YAML 时的内置主代理（规划/重规划侧）提示。
@@ -122,7 +123,9 @@ func DefaultPlanExecuteOrchestratorInstruction() string {
 
 ## 表达
 
-在调用工具或给出计划变更前，用 2～5 句中文说明当前决策依据与期望证据形态；最终对用户交付结构化结论（发现摘要、证据、风险、下一步）。`
+在调用工具或给出计划变更前，用 2～5 句中文说明当前决策依据与期望证据形态；最终对用户交付结构化结论（发现摘要、证据、风险、下一步）。
+
+` + projectprompt.ShellExecExecuteGuidanceSection()
 }
 
 // DefaultSupervisorOrchestratorInstruction 当未配置 supervisor 专用 Markdown / YAML 时的内置监督者提示（transfer / exit 说明仍由运行时在末尾追加）。

internal/multiagent/orphan_tool_pruner_middleware.go

@@ -27,7 +27,7 @@ import (
 //     本中间件与之互补，专职兜底正向孤儿。
 //   - 仅剔除消息，不向历史里注入虚构 assistant(tc)：虚构 tool_calls 反而会误导模型后续推理。
 //     摘要已覆盖被裁剪段的语义，丢一条原始 tool 结果对对话连贯性影响最小。
-//   - 位置建议：挂在所有可能改写历史的中间件（summarization / reduction / skill / plantask /
+//   - 位置建议：挂在 summarization / reduction / skill / plantask / system 合并 / 续聊 dedup 之后，
 //     tool_search）之后，靠近 ChatModel 调用的那一端。
 type orphanToolPrunerMiddleware struct {
 	adk.BaseChatModelAgentMiddleware

internal/multiagent/runner.go

@@ -20,6 +20,7 @@ import (
 	"cyberstrike-ai/internal/openai"
 	"cyberstrike-ai/internal/project"
 	"cyberstrike-ai/internal/reasoning"
+	"cyberstrike-ai/internal/security"
 
 	einoopenai "github.com/cloudwego/eino-ext/components/model/openai"
 	"github.com/cloudwego/eino/adk"
@@ -120,7 +121,7 @@ func RunDeepAgent(
 		mcpIDs = append(mcpIDs, id)
 		mcpIDsMu.Unlock()
 	}
-	einoExecMonitor := newEinoExecuteMonitorCallback(ag, recorder)
+	einoExecBegin, einoExecFinish := newEinoExecuteMonitorCallbacks(ag, recorder)
 
 	// 与单代理流式一致：在 response_start / response_delta 的 data 中带当前 mcpExecutionIds，供主聊天绑定复制与展示。
 	snapshotMCPIDs := func() []string {
@@ -223,7 +224,7 @@ func RunDeepAgent(
 			}
 			if einoSkillMW != nil {
 				if einoFSTools && einoLoc != nil {
-					subFs, fsErr := subAgentFilesystemMiddleware(ctx, einoLoc, toolInvokeNotify, id, einoExecMonitor, agentToolTimeoutMinutes(appCfg), nil)
+					subFs, fsErr := subAgentFilesystemMiddleware(ctx, einoLoc, toolInvokeNotify, id, einoExecBegin, einoExecFinish, agentToolTimeoutMinutes(appCfg), agentShellNoOutputTimeoutSeconds(appCfg), nil)
 					if fsErr != nil {
 						return nil, fmt.Errorf("子代理 %q filesystem 中间件: %w", id, fsErr)
 					}
@@ -231,13 +232,13 @@ func RunDeepAgent(
 				}
 				subHandlers = append(subHandlers, einoSkillMW)
 			}
-			subHandlers = append(subHandlers, subSumMw)
-			// 孤儿 tool 消息兜底：放在 summarization 之后，telemetry 之前，
-			// 以便 telemetry 记录的 token 数与 LLM 实际入参一致。
-			subHandlers = append(subHandlers, newOrphanToolPrunerMiddleware(logger, "sub_agent:"+id))
-			if teleMw := newEinoModelInputTelemetryMiddleware(logger, appCfg.OpenAI.Model, conversationID, "sub_agent"); teleMw != nil {
-				subHandlers = append(subHandlers, teleMw)
-			}
+			subHandlers = appendEinoChatModelTailMiddlewares(subHandlers, einoChatModelTailConfig{
+				logger:         logger,
+				phase:          "sub_agent:" + id,
+				summarization:  subSumMw,
+				modelName:      appCfg.OpenAI.Model,
+				conversationID: conversationID,
+			})
 
 			subInstrFinal := project.AppendVisionImageAnalysisIfReady(instr, appCfg.Vision.Ready())
 			subInstrFinal = injectToolNamesOnlyInstruction(ctx, subInstrFinal, subTools, subToolSearchActive)
@@ -253,10 +254,11 @@ func RunDeepAgent(
 				)
 			}
 			sa, err := adk.NewChatModelAgent(ctx, &adk.ChatModelAgentConfig{
-				Name:        id,
-				Description: desc,
-				Instruction: subInstrFinal,
-				Model:       subModel,
+				Name:          id,
+				Description:   desc,
+				Instruction:   subInstrFinal,
+				GenModelInput: literalInstructionGenModelInput,
+				Model:         subModel,
 				ToolsConfig: adk.ToolsConfig{
 					ToolsNodeConfig: compose.ToolsNodeConfig{
 						Tools:               subToolsForCfg,
@@ -358,19 +360,28 @@ func RunDeepAgent(
 	if einoLoc != nil && einoFSTools {
 		deepBackend = einoLoc
 		deepShell = &einoStreamingShellWrap{
-			inner:              einoLoc,
-			invokeNotify:       toolInvokeNotify,
-			einoAgentName:      orchestratorName,
-			outputChunk:        nil,
-			recordMonitor:      einoExecMonitor,
-			toolTimeoutMinutes: agentToolTimeoutMinutes(appCfg),
+			inner:                   security.NewEinoStreamingShell(),
+			invokeNotify:            toolInvokeNotify,
+			einoAgentName:           orchestratorName,
+			outputChunk:             nil,
+			beginMonitor:            einoExecBegin,
+			finishMonitor:           einoExecFinish,
+			toolTimeoutMinutes:      agentToolTimeoutMinutes(appCfg),
+			shellNoOutputTimeoutSec: agentShellNoOutputTimeoutSeconds(appCfg),
 		}
 	}
 
 	// noNestedTaskMiddleware 必须在最外层（最先拦截），防止 skill 或其他中间件内部触发 task 调用绕过检测。
 	deepHandlers := []adk.ChatModelAgentMiddleware{newNoNestedTaskMiddleware()}
-	taskEnrichExtra := systemPromptExtra
-	if mw := newTaskContextEnrichMiddleware(userMessage, history, ma.SubAgentUserContextMaxRunes, taskEnrichExtra); mw != nil {
+	var taskBlackboardSupplement string
+	if appCfg.Project.Enabled && db != nil {
+		if pid := strings.TrimSpace(projectID); pid != "" {
+			if block, err := project.BuildFactIndexBlock(db, pid, appCfg.Project); err == nil {
+				taskBlackboardSupplement = strings.TrimSpace(block)
+			}
+		}
+	}
+	if mw := newTaskContextEnrichMiddleware(userMessage, history, ma.SubAgentUserContextMaxRunesEffective(), taskBlackboardSupplement); mw != nil {
 		deepHandlers = append(deepHandlers, mw)
 	}
 	if len(mainOrchestratorPre) > 0 {
@@ -379,14 +390,14 @@ func RunDeepAgent(
 	if einoSkillMW != nil {
 		deepHandlers = append(deepHandlers, einoSkillMW)
 	}
-	deepHandlers = append(deepHandlers, mainSumMw)
-	deepHandlers = append(deepHandlers, newOrphanToolPrunerMiddleware(logger, "deep_orchestrator"))
-	if teleMw := newEinoModelInputTelemetryMiddleware(logger, appCfg.OpenAI.Model, conversationID, "deep_orchestrator"); teleMw != nil {
-		deepHandlers = append(deepHandlers, teleMw)
-	}
-	if capMw := newModelFacingTraceMiddleware(modelFacingTrace); capMw != nil {
-		deepHandlers = append(deepHandlers, capMw)
-	}
+	deepHandlers = appendEinoChatModelTailMiddlewares(deepHandlers, einoChatModelTailConfig{
+		logger:         logger,
+		phase:          "deep_orchestrator",
+		summarization:  mainSumMw,
+		modelName:      appCfg.OpenAI.Model,
+		conversationID: conversationID,
+		trace:          modelFacingTrace,
+	})
 
 	supHandlers := []adk.ChatModelAgentMiddleware{}
 	if len(mainOrchestratorPre) > 0 {
@@ -395,14 +406,14 @@ func RunDeepAgent(
 	if einoSkillMW != nil {
 		supHandlers = append(supHandlers, einoSkillMW)
 	}
-	supHandlers = append(supHandlers, mainSumMw)
-	supHandlers = append(supHandlers, newOrphanToolPrunerMiddleware(logger, "supervisor_orchestrator"))
-	if teleMw := newEinoModelInputTelemetryMiddleware(logger, appCfg.OpenAI.Model, conversationID, "supervisor_orchestrator"); teleMw != nil {
-		supHandlers = append(supHandlers, teleMw)
-	}
-	if capMw := newModelFacingTraceMiddleware(modelFacingTrace); capMw != nil {
-		supHandlers = append(supHandlers, capMw)
-	}
+	supHandlers = appendEinoChatModelTailMiddlewares(supHandlers, einoChatModelTailConfig{
+		logger:         logger,
+		phase:          "supervisor_orchestrator",
+		summarization:  mainSumMw,
+		modelName:      appCfg.OpenAI.Model,
+		conversationID: conversationID,
+		trace:          modelFacingTrace,
+	})
 
 	mainToolsCfg := adk.ToolsConfig{
 		ToolsNodeConfig: compose.ToolsNodeConfig{
@@ -416,11 +427,27 @@ func RunDeepAgent(
 		EmitInternalEvents: true,
 	}
 
-	deepOutKey, modelRetry, taskGen := deepExtrasFromConfig(ma)
+	deepOutKey, taskGen := deepExtrasFromConfig(ma)
 
 	var da adk.Agent
 	switch orchMode {
 	case "plan_execute":
+		plannerModelCfg := &einoopenai.ChatModelConfig{
+			APIKey:     appCfg.OpenAI.APIKey,
+			BaseURL:    strings.TrimSuffix(appCfg.OpenAI.BaseURL, "/"),
+			Model:      appCfg.OpenAI.Model,
+			HTTPClient: httpClient,
+		}
+		reasoning.ApplyPlanExecutePlannerModelConfig(plannerModelCfg, &appCfg.OpenAI)
+		peMainModel, perr := einoopenai.NewChatModel(ctx, plannerModelCfg)
+		if perr != nil {
+			return nil, fmt.Errorf("plan_execute 规划模型: %w", perr)
+		}
+		if logger != nil {
+			logger.Info("plan_execute: planner/replanner 使用无 reasoning 的独立 ChatModel（ToolChoiceForced 兼容）",
+				zap.String("model", appCfg.OpenAI.Model),
+			)
+		}
 		execModel, perr := einoopenai.NewChatModel(ctx, baseModelCfg)
 		if perr != nil {
 			return nil, fmt.Errorf("plan_execute 执行器模型: %w", perr)
@@ -428,13 +455,13 @@ func RunDeepAgent(
 		// 构建 filesystem 中间件（与 Deep sub-agent 一致）
 		var peFsMw adk.ChatModelAgentMiddleware
 		if einoSkillMW != nil && einoFSTools && einoLoc != nil {
-			peFsMw, err = subAgentFilesystemMiddleware(ctx, einoLoc, toolInvokeNotify, "executor", einoExecMonitor, agentToolTimeoutMinutes(appCfg), nil)
+			peFsMw, err = subAgentFilesystemMiddleware(ctx, einoLoc, toolInvokeNotify, "executor", einoExecBegin, einoExecFinish, agentToolTimeoutMinutes(appCfg), agentShellNoOutputTimeoutSeconds(appCfg), nil)
 			if err != nil {
 				return nil, fmt.Errorf("plan_execute filesystem 中间件: %w", err)
 			}
 		}
 		peRoot, perr := NewPlanExecuteRoot(ctx, &PlanExecuteRootArgs{
-			MainToolCallingModel: mainModel,
+			MainToolCallingModel: peMainModel,
 			ExecModel:            execModel,
 			OrchInstruction:      orchInstruction,
 			ToolsCfg:             mainToolsCfg,
@@ -451,12 +478,14 @@ func RunDeepAgent(
 			SkillMiddleware:      einoSkillMW,
 			FilesystemMiddleware: peFsMw,
 			ModelFacingTrace:     modelFacingTrace,
-			PlannerReplannerRewriteHandlers: []adk.ChatModelAgentMiddleware{
-				mainSumMw,
-				// 孤儿 tool 消息兜底：必须挂在 summarization 之后、telemetry 之前。
-				newOrphanToolPrunerMiddleware(logger, "plan_execute_planner_replanner"),
-				newEinoModelInputTelemetryMiddleware(logger, appCfg.OpenAI.Model, conversationID, "plan_execute_planner_replanner_rewrite"),
-			},
+			PlannerReplannerRewriteHandlers: appendEinoChatModelTailMiddlewares(nil, einoChatModelTailConfig{
+				logger:         logger,
+				phase:          "plan_execute_planner_replanner",
+				summarization:  mainSumMw,
+				modelName:      appCfg.OpenAI.Model,
+				conversationID: conversationID,
+				skipTrace:      true,
+			}),
 		})
 		if perr != nil {
 			return nil, perr
@@ -467,15 +496,13 @@ func RunDeepAgent(
 			Name:          orchestratorName,
 			Description:   orchDescription,
 			Instruction:   supInstr,
+			GenModelInput: literalInstructionGenModelInput,
 			Model:         mainModel,
 			ToolsConfig:   mainToolsCfg,
 			MaxIterations: deepMaxIter,
 			Handlers:      supHandlers,
 			Exit:          &adk.ExitTool{},
 		}
-		if modelRetry != nil {
-			supCfg.ModelRetryConfig = modelRetry
-		}
 		if deepOutKey != "" {
 			supCfg.OutputKey = deepOutKey
 		}
@@ -509,9 +536,6 @@ func RunDeepAgent(
 		if deepOutKey != "" {
 			dcfg.OutputKey = deepOutKey
 		}
-		if modelRetry != nil {
-			dcfg.ModelRetryConfig = modelRetry
-		}
 		if taskGen != nil {
 			dcfg.TaskToolDescriptionGenerator = taskGen
 		}

internal/multiagent/shell_tool_guidance.go

@@ -0,0 +1,33 @@
+package multiagent
+
+import (
+	"strings"
+
+	"cyberstrike-ai/internal/projectprompt"
+)
+
+func shellToolsPresent(toolNames []string) bool {
+	for _, n := range toolNames {
+		switch strings.ToLower(strings.TrimSpace(n)) {
+		case "exec", "execute":
+			return true
+		}
+	}
+	return false
+}
+
+// injectShellToolGuidance 在系统提示末尾追加 exec/execute 分工（仅当工具列表含 exec 或 execute）。
+func injectShellToolGuidance(instruction string, toolNames []string) string {
+	if !shellToolsPresent(toolNames) {
+		return instruction
+	}
+	block := strings.TrimSpace(projectprompt.ShellExecExecuteGuidanceSection())
+	if block == "" {
+		return instruction
+	}
+	s := strings.TrimSpace(instruction)
+	if s == "" {
+		return block
+	}
+	return s + "\n\n" + block
+}

internal/multiagent/shell_tool_guidance_test.go

@@ -0,0 +1,17 @@
+package multiagent
+
+import (
+	"strings"
+	"testing"
+)
+
+func TestInjectShellToolGuidance(t *testing.T) {
+	got := injectShellToolGuidance("base", []string{"nmap"})
+	if got != "base" {
+		t.Fatalf("expected unchanged, got %q", got)
+	}
+	got = injectShellToolGuidance("base", []string{"exec", "nmap"})
+	if !strings.Contains(got, "exec/execute") || !strings.Contains(got, "base") {
+		t.Fatalf("expected shell guidance appended, got %q", got)
+	}
+}

internal/multiagent/sub_agent_context.go

@@ -3,6 +3,7 @@ package multiagent
 import (
 	"context"
 	"encoding/json"
+	"fmt"
 	"strings"
 
 	"cyberstrike-ai/internal/agent"
@@ -11,7 +12,7 @@ import (
 	"github.com/cloudwego/eino/components/tool"
 )
 
-const defaultSubAgentUserContextMaxRunes = 2000
+const userContextSupplementHeader = "\n\n## 用户历史输入（原文，子代理必读）\n"
 
 // taskContextEnrichMiddleware intercepts "task" tool calls on the orchestrator
 // and appends the user's original conversation messages to the task description.
@@ -30,13 +31,14 @@ type taskContextEnrichMiddleware struct {
 // newTaskContextEnrichMiddleware returns a middleware that enriches task
 // descriptions with user conversation context. Returns nil if disabled
 // (maxRunes < 0) or no user messages exist.
+// projectBlackboard 仅传项目黑板索引块（BuildFactIndexBlock）；勿传完整 systemPromptExtra。
 func newTaskContextEnrichMiddleware(userMessage string, history []agent.ChatMessage, maxRunes int, projectBlackboard string) adk.ChatModelAgentMiddleware {
 	supplement := buildUserContextSupplement(userMessage, history, maxRunes)
 	if bb := strings.TrimSpace(projectBlackboard); bb != "" {
 		if supplement != "" {
-			supplement += "\n\n## 项目黑板索引\n" + bb
+			supplement += "\n\n" + bb
 		} else {
-			supplement = "\n\n## 项目黑板索引\n" + bb
+			supplement = "\n\n" + bb
 		}
 	}
 	if supplement == "" {
@@ -86,9 +88,6 @@ func buildUserContextSupplement(userMessage string, history []agent.ChatMessage,
 	if maxRunes < 0 {
 		return ""
 	}
-	if maxRunes == 0 {
-		maxRunes = defaultSubAgentUserContextMaxRunes
-	}
 
 	var userMsgs []string
 	for _, h := range history {
@@ -107,12 +106,16 @@ func buildUserContextSupplement(userMessage string, history []agent.ChatMessage,
 		return ""
 	}
 
-	joined := strings.Join(userMsgs, "\n---\n")
-	if len([]rune(joined)) > maxRunes {
+	lines := make([]string, 0, len(userMsgs))
+	for i, msg := range userMsgs {
+		lines = append(lines, fmt.Sprintf("[第%d轮] %s", i+1, msg))
+	}
+	joined := strings.Join(lines, "\n")
+	if maxRunes > 0 && len([]rune(joined)) > maxRunes {
 		joined = truncateKeepFirstLast(userMsgs, maxRunes)
 	}
 
-	return "\n\n## 会话上下文（自动补充，确保你了解用户完整意图）\n" + joined
+	return userContextSupplementHeader + joined
 }
 
 // truncateKeepFirstLast keeps the first and last user messages, giving each

internal/multiagent/sub_agent_context_test.go

@@ -74,7 +74,7 @@ func TestBuildUserContextSupplement_DisabledByNegative(t *testing.T) {
 func TestBuildUserContextSupplement_CustomMaxRunes(t *testing.T) {
 	msg := strings.Repeat("A", 200)
 	result := buildUserContextSupplement(msg, nil, 50)
-	header := "\n\n## 会话上下文（自动补充，确保你了解用户完整意图）\n"
+	header := userContextSupplementHeader
 	body := strings.TrimPrefix(result, header)
 	if len([]rune(body)) > 50 {
 		t.Errorf("body should be capped at 50 runes, got %d", len([]rune(body)))
@@ -89,7 +89,7 @@ func TestBuildUserContextSupplement_TruncateKeepsFirstAndLast(t *testing.T) {
 		history = append(history, agent.ChatMessage{Role: "user", Content: strings.Repeat("B", 500)})
 	}
 	last := "最后一条指令"
-	result := buildUserContextSupplement(last, history, 0)
+	result := buildUserContextSupplement(last, history, 800)
 	if !strings.Contains(result, "http://target.com") {
 		t.Error("first message (target URL) should survive truncation")
 	}