docs: sync to Claude Code v2.1.245 (#177)

* docs: sync to Claude Code v2.1.245 Sync the tutorial, its four translations, and both learner-facing skills against Claude Code v2.1.245 (was anchored at v2.1.235). The graded material turned out to be staler than the tutorials it grades. The lesson READMEs already said "31 hook events", "six rewind options" and "six permission modes"; the quiz and self-assessment still said 25, 5 and 5, so learners were scored against a wrong answer and then sent to a lesson that contradicted it. Seven such claims are fixed here. Two long-standing doc defects were also verified against official docs: - Hook default timeout was documented as 60 seconds. It is 600 for command/http/mcp_tool (30 for prompt, 60 for agent), with per-event overrides. Wrong in five places. - Skill precedence was documented as enterprise > project > personal. The official order is enterprise > personal > project. question-bank.md Lesson 03 Q9 already graded the correct order, so the lesson had been contradicting its own quiz. Fixed - Hook default timeout 60s -> 600s, with per-event overrides - Skill precedence: project and personal were swapped - Memory "overrides root CLAUDE.md" -> concatenation model (last holdouts after the quiz was aligned in #164) - permissionMode frontmatter omitted `auto` (English + vi/ja/uk); the quiz was right and the docs were wrong - Built-in subagent `Bash` no longer exists upstream -> renamed to `claude` - `/undo` is absent from the official commands reference -> marked deprecated - `backgroundTasks` settings block was fabricated -> replaced with CLAUDE_CODE_DISABLE_BACKGROUND_TASKS and the concurrency env var - Plugin manifest field is `lspServers`, not inline `lsp` - vi/05-mcp documented `claude mcp add --transport ws`, which upstream explicitly does not support - Self-assessment could never reach its own maximum score: rounds cap at 4+4+4+3+4 = 19, not 20 - 13 dangling quiz `Review:` pointers, two of which named another lesson - Skill description cap: 1024 chars -> 1,536 combined with `when_to_use` Added - WebSocket (`ws`) MCP transport, undocumented until now - Concise output style (v2.1.237), a fifth built-in - modelPicker, promptCacheTtl, subagentPromptCacheTtl, modelPricing, keybindingFlavor settings - ANTHROPIC_DEFAULT_MODEL env var - Hook `async`, `asyncRewake`, `shell`, `statusMessage`, `retry` keys; `agent` type marked experimental - /design canvas, Remote Control out of research preview, notify_when_idle, plugin manifest fields (workflows, channels, dependencies, outputStyles) Deferred deliberately: the reserved `synced` skills folder, and a bulk footer stamp across the translations. zh/ carries none of the P0 claims (it is abridged upstream of this run), so it has no P0 sub-tasks rather than fabricated English. Mermaid validation did not run in this environment (mmdc absent); CI covers it. * fix(docs): propagate v2.1.245 corrections to CATALOG, translations, and stale enumerations * fix(docs): propagate built-in subagent rename to uk/vi/zh CATALOG and zh lesson
2026-08-30 12:30:36 +02:00 · 2026-08-26 07:20:04 +02:00
parent 1c04dbf88e
commit 28149a85fc
67 changed files with 754 additions and 386 deletions
@@ -3,7 +3,7 @@ name: lesson-quiz
 description: "Test a learner on a single Claude Code tutorial lesson (01-10) with 10 questions, scoring answers and flagging weak spots. Use before, during, or after a lesson. Don't use for whole-tutorial assessment or explaining a topic instead of testing it."
 effort: high
 metadata:
-  version: 1.1.0
+  version: 1.2.0
  author: Luong NGUYEN
 ---

@@ -12,7 +12,7 @@
 - **Options**: A) Built-in, skills, plugin commands, MCP prompts | B) Built-in, custom, hook commands, API prompts | C) System, user, plugin, terminal commands | D) Core, extension, macro, script commands
 - **Correct**: A
 - **Explanation**: Claude Code has built-in commands (like /help, /compact), skills (SKILL.md files), plugin commands (namespaced plugin-name:command), and MCP prompts (/mcp__server__prompt).
- **Review**: Types of Slash Commands section
+- **Review**: Overview section

 ### Q2
 - **Category**: practical
@@ -28,7 +28,7 @@
 - **Options**: A) The legacy command | B) The skill | C) Whichever was created first | D) Claude asks the user to choose
 - **Correct**: B
 - **Explanation**: Skills take precedence over legacy commands with the same name. The skill system supersedes the older command system.
- **Review**: Skill precedence section
+- **Review**: Skill vs Command Conflict section

 ### Q4
 - **Category**: practical
@@ -44,7 +44,7 @@
 - **Options**: A) Prevents the skill from running entirely | B) Allows only the user to invoke it (Claude cannot auto-invoke) | C) Hides it from the /help menu | D) Disables the skill's AI processing
 - **Correct**: B
 - **Explanation**: `disable-model-invocation: true` means only the user can trigger the command via `/command-name`. Claude will never auto-invoke it, useful for skills with side effects like deployments.
- **Review**: Controlling invocation section
+- **Review**: Frontmatter Reference section

 ### Q6
 - **Category**: practical
@@ -52,7 +52,7 @@
 - **Options**: A) `disable-model-invocation: true` | B) `user-invocable: false` | C) `hidden: true` | D) `auto-only: true`
 - **Correct**: B
 - **Explanation**: `user-invocable: false` hides the skill from the user's slash menu but allows Claude to invoke it automatically based on context.
- **Review**: Invocation control matrix
+- **Review**: Frontmatter Reference section

 ### Q7
 - **Category**: practical
@@ -60,7 +60,7 @@
 - **Options**: A) `.claude/commands/deploy.md` | B) `.claude/skills/deploy/SKILL.md` | C) `.claude/skills/deploy.md` | D) `.claude/deploy/SKILL.md`
 - **Correct**: B
 - **Explanation**: Skills live in a directory under `.claude/skills/` with a `SKILL.md` file inside. The directory name matches the command name.
- **Review**: Skill types and locations section
+- **Review**: Creating a Custom Command as a Skill section

 ### Q8
 - **Category**: conceptual
@@ -112,7 +112,7 @@
 - **Options**: A) 3 levels deep | B) 4 levels deep | C) 10 levels deep | D) Unlimited
 - **Correct**: B
 - **Explanation**: The `@path` import syntax supports recursive imports up to a maximum depth of 4 hops to prevent infinite loops.
- **Review**: Import syntax section
+- **Review**: Memory Architecture section

 ### Q4
 - **Category**: practical
@@ -127,7 +127,7 @@
 - **Question**: How many lines of Auto Memory's MEMORY.md are loaded at session start?
 - **Options**: A) All lines | B) First 100 lines | C) First 200 lines | D) First 500 lines
 - **Correct**: C
- **Explanation**: The first 200 lines of MEMORY.md are auto-loaded into context at session start. Topic files referenced from MEMORY.md are loaded on demand.
+- **Explanation**: The first 200 lines of MEMORY.md — or the first 25 KB, whichever comes first — are auto-loaded into context at session start. Topic files referenced from MEMORY.md are loaded on demand.
 - **Review**: Auto Memory section

 ### Q6
@@ -188,7 +188,7 @@
 - **Options**: A) The skill's file name | B) The `description` field in frontmatter with when-to-use keywords | C) The skill's directory location | D) The `auto-invoke: true` frontmatter field
 - **Correct**: B
 - **Explanation**: Claude decides whether to auto-invoke a skill based solely on its `description` field. It must include specific trigger phrases and scenarios.
- **Review**: Auto-invocation section
+- **Review**: Controlling Skill Invocation section

 ### Q3
 - **Category**: conceptual
@@ -196,7 +196,7 @@
 - **Options**: A) 100 lines | B) 250 lines | C) 500 lines | D) 1000 lines
 - **Correct**: C
 - **Explanation**: SKILL.md should be kept under 500 lines. Larger reference material belongs in `references/` subdirectory files.
- **Review**: Content guidelines section
+- **Review**: Keep SKILL.md Under 500 Lines section

 ### Q4
 - **Category**: practical
@@ -212,7 +212,7 @@
 - **Options**: A) 0.5% of context window | B) 1% of context window | C) 5% of context window | D) 10% of context window
 - **Correct**: B
 - **Explanation**: Skill metadata occupies about 1% of the context window (fallback: 8,000 characters). This is configurable with `SLASH_COMMAND_TOOL_CHAR_BUDGET`.
- **Review**: Context budget section
+- **Review**: the "Description budget" paragraph under Automatic Discovery

 ### Q6
 - **Category**: practical
@@ -276,10 +276,10 @@

 ### Q3
 - **Category**: conceptual
- **Question**: Which built-in subagent uses the Haiku model and is optimized for read-only codebase exploration?
+- **Question**: Which built-in subagent is optimized for read-only codebase exploration?
 - **Options**: A) general-purpose | B) Plan | C) Explore | D) Bash
 - **Correct**: C
- **Explanation**: The Explore subagent uses Haiku for fast, read-only codebase exploration. It supports three thoroughness levels: quick, medium, very thorough.
+- **Explanation**: The Explore subagent is the read-only codebase search agent. Since v2.1.198 it inherits the session model, capped at Opus — `model: haiku` is opt-in, not the default. It supports three thoroughness levels: quick, medium, very thorough.
 - **Review**: Built-in subagents section

 ### Q4
@@ -300,10 +300,10 @@

 ### Q6
 - **Category**: practical
- **Question**: How do you make a subagent run in the background?
+- **Question**: How do you force a subagent to always run in the background (never inline)?
 - **Options**: A) Set `background: true` in the agent config | B) Use `async: true` in the agent config | C) Press Ctrl+D after starting it | D) Use `--background` CLI flag
 - **Correct**: A
- **Explanation**: `background: true` in the agent configuration makes the subagent always run as a background task. Users can also use Ctrl+B to send a foreground task to background.
+- **Explanation**: `background: true` in the agent configuration makes the subagent always run as a background task and prevents it from running inline. Subagents already run in the background by default since v2.1.198; this field removes the inline option. Users can also use Ctrl+B to send a foreground task to background.
 - **Review**: Background subagents section

 ### Q7
@@ -325,9 +325,9 @@
 ### Q9
 - **Category**: conceptual
 - **Question**: What are the valid `permissionMode` values for a subagent?
- **Options**: A) read, write, admin | B) default, acceptEdits, bypassPermissions, plan, dontAsk, auto | C) safe, normal, dangerous | D) restricted, standard, elevated
+- **Options**: A) read, write, admin | B) manual, acceptEdits, bypassPermissions, plan, dontAsk, auto | C) safe, normal, dangerous | D) restricted, standard, elevated
 - **Correct**: B
- **Explanation**: Subagents support six permission modes: default (prompts for everything), acceptEdits (auto-accepts file edits), bypassPermissions (skips all), plan (read-only), dontAsk (auto-denies unless pre-approved), auto (background classifier decides).
+- **Explanation**: Subagents support six permission modes: manual (prompts for everything — renamed from `default` in v2.1.200, which is still accepted as an alias), acceptEdits (auto-accepts file edits), bypassPermissions (skips all), plan (read-only), dontAsk (auto-denies unless pre-approved), auto (background classifier decides).
 - **Review**: Configuration fields section

 ### Q10
@@ -344,11 +344,11 @@

 ### Q1
 - **Category**: conceptual
- **Question**: What are the three MCP transport protocols, and which is recommended?
- **Options**: A) HTTP (recommended), Stdio, SSE (deprecated) | B) WebSocket (recommended), REST, gRPC | C) TCP, UDP, HTTP | D) Stdio (recommended), HTTP, SSE
+- **Question**: What are the four MCP transport protocols, and which is recommended?
+- **Options**: A) HTTP (recommended), Stdio, SSE (deprecated), WebSocket (`ws`) | B) REST (recommended), GraphQL, gRPC, Thrift | C) TCP, UDP, HTTP | D) Stdio (recommended), HTTP, SSE
 - **Correct**: A
- **Explanation**: HTTP is recommended for remote servers. Stdio is for local processes (most common currently). SSE is deprecated but still supported.
- **Review**: Transport protocols section
+- **Explanation**: HTTP is recommended for remote servers. Stdio is for local processes (most common currently). SSE is deprecated but still supported. WebSocket (`type: "ws"`) holds a persistent bidirectional connection for servers that push events unprompted; it is added with `claude mcp add-json` or `.mcp.json`, not with `--transport`, and supports header-only auth.
+- **Review**: MCP Installation Methods section

 ### Q2
 - **Category**: practical
@@ -428,10 +428,10 @@

 ### Q1
 - **Category**: conceptual
- **Question**: What are the four types of hooks in Claude Code?
- **Options**: A) Pre, Post, Error, and Filter hooks | B) Command, HTTP, Prompt, and Agent hooks | C) Before, After, Around, and Through hooks | D) Input, Output, Filter, and Transform hooks
+- **Question**: What are the five types of hooks in Claude Code?
+- **Options**: A) Pre, Post, Error, Filter, and Transform hooks | B) Command, HTTP, MCP Tool, Prompt, and Agent hooks | C) Before, After, Around, Through, and Beside hooks | D) Input, Output, Filter, Transform, and Validate hooks
 - **Correct**: B
- **Explanation**: Command hooks run shell scripts, HTTP hooks call webhook endpoints, Prompt hooks use single-turn LLM evaluation, and Agent hooks use subagent-based verification.
+- **Explanation**: Command hooks run shell scripts, HTTP hooks call webhook endpoints, MCP Tool hooks invoke an MCP tool, Prompt hooks use single-turn LLM evaluation, and Agent hooks use subagent-based verification. Agent hooks are experimental and may change.
 - **Review**: Hook types section

 ### Q2
@@ -463,7 +463,7 @@
 - **Question**: Which hook event supports `CLAUDE_ENV_FILE` for persisting environment variables into the session?
 - **Options**: A) PreToolUse | B) UserPromptSubmit | C) SessionStart | D) All events
 - **Correct**: C
- **Explanation**: Only SessionStart hooks can use `CLAUDE_ENV_FILE` to persist environment variables into the session.
+- **Explanation**: `CLAUDE_ENV_FILE` is available to SessionStart, CwdChanged, and FileChanged hooks, which use it to persist environment variables into the session. Of the options listed, SessionStart is the one that supports it — PreToolUse and UserPromptSubmit do not, and it is not available to all events.
 - **Review**: SessionStart section

 ### Q6
@@ -493,9 +493,9 @@
 ### Q9
 - **Category**: conceptual
 - **Question**: How many hook events does Claude Code support in total?
- **Options**: A) 10 | B) 16 | C) 25 | D) 30
+- **Options**: A) 16 | B) 25 | C) 31 | D) 40
 - **Correct**: C
- **Explanation**: Claude Code supports 25 hook events: PreToolUse, PostToolUse, PostToolUseFailure, UserPromptSubmit, Stop, StopFailure, SubagentStop, SubagentStart, PermissionRequest, Notification, PreCompact, PostCompact, SessionStart, SessionEnd, WorktreeCreate, WorktreeRemove, ConfigChange, CwdChanged, FileChanged, TeammateIdle, TaskCompleted, TaskCreated, Elicitation, ElicitationResult, InstructionsLoaded.
+- **Explanation**: Claude Code supports 31 hook events. They fall into four groups: session lifecycle (`SessionStart`, `Setup`, `SessionEnd`), prompt and message handling (`UserPromptSubmit`, `UserPromptExpansion`, `MessageDisplay`), tool use and permissions (`PreToolUse`, `PostToolUse`, `PermissionRequest`, ...), and agent, task, and environment coordination (`SubagentStart`, `TaskCreated`, `CwdChanged`, `WorktreeCreate`, ...). See the full event table in the lesson.
 - **Review**: Hook events table

 ### Q10
@@ -532,7 +532,7 @@
 - **Options**: A) `$PLUGIN_HOME` | B) `${CLAUDE_PLUGIN_ROOT}` | C) `$PLUGIN_DIR` | D) `${CLAUDE_PLUGIN_PATH}`
 - **Correct**: B
 - **Explanation**: `${CLAUDE_PLUGIN_ROOT}` resolves to the plugin's installed directory, enabling portable path references in hooks and MCP configs.
- **Review**: Plugin directory structure section
+- **Review**: Plugin Structure Example section

 ### Q4
 - **Category**: practical
@@ -540,15 +540,15 @@
 - **Options**: A) `/check-security` | B) `/pr-review:check-security` | C) `/plugin pr-review check-security` | D) `/pr-review/check-security`
 - **Correct**: B
 - **Explanation**: Plugin commands use a `plugin-name:command-name` namespace to avoid conflicts with user commands and other plugins.
- **Review**: Plugin commands section
+- **Review**: Plugin Commands section in Lesson 01 (Slash Commands)

 ### Q5
 - **Category**: conceptual
 - **Question**: Which components can a plugin bundle?
- **Options**: A) Only commands and settings | B) Commands, agents, skills, hooks, MCP servers, LSP config, settings, templates, scripts | C) Only commands, hooks, and MCP servers | D) Only skills and agents
+- **Options**: A) Only commands and settings | B) Skills, agents, commands, hooks, MCP servers, LSP config, settings, templates, scripts | C) Only commands, hooks, and MCP servers | D) Only skills and agents
 - **Correct**: B
- **Explanation**: Plugins can bundle: commands/, agents/, skills/, hooks/hooks.json, .mcp.json, .lsp.json, settings.json, templates/, scripts/, docs/, tests/.
- **Review**: Plugin directory structure section
+- **Explanation**: Plugins can bundle: skills/, agents/, commands/, hooks/hooks.json, .mcp.json, .lsp.json, settings.json, templates/, scripts/, docs/, tests/. Note that `commands/` is legacy — official guidance is to use `skills/` for new plugins, though existing `commands/` directories keep working.
+- **Review**: Plugin Structure Example section

 ### Q6
 - **Category**: practical
@@ -605,7 +605,7 @@
 ### Q2
 - **Category**: practical
 - **Question**: How do you access the checkpoint browser?
- **Options**: A) Use `/checkpoints` command | B) Press `Esc + Esc` (double-escape) or use `/rewind` | C) Use `/history` command | D) Press `Ctrl+Z`
+- **Options**: A) Use `/restore` command | B) Press `Esc + Esc` (double-escape) or use `/rewind` | C) Use `/history` command | D) Press `Ctrl+Z`
 - **Correct**: B
 - **Explanation**: Double-escape (Esc+Esc) or the `/rewind` command opens the checkpoint browser to select a restore point.
 - **Review**: Accessing checkpoints section
@@ -631,8 +631,8 @@
 - **Question**: How long are checkpoints retained?
 - **Options**: A) Until session ends | B) 7 days | C) 30 days | D) Indefinitely
 - **Correct**: C
- **Explanation**: Checkpoints persist across sessions for up to 30 days, after which they are automatically cleaned up.
- **Review**: Checkpoint persistence section
+- **Explanation**: Checkpoints persist across sessions for up to 30 days, after which they are automatically cleaned up. Retention is also capped by count: only the 100 most recent checkpoints per session are kept, so older ones are dropped even inside the 30-day window. The window is governed by `cleanupPeriodDays`.
+- **Review**: Checkpoint Retention section

 ### Q6
 - **Category**: practical
@@ -640,7 +640,7 @@
 - **Options**: A) Deletes the conversation from that point | B) Compresses the conversation into an AI-generated summary while preserving the original in the transcript | C) Creates a bullet-point list of changes | D) Exports the conversation to a file
 - **Correct**: B
 - **Explanation**: Summarize compresses the conversation into a shorter AI-generated summary. The original full text is preserved in the transcript file.
- **Review**: Summarize option section
+- **Review**: Rewind Options section

 ### Q7
 - **Category**: conceptual
@@ -681,9 +681,9 @@
 ### Q1
 - **Category**: conceptual
 - **Question**: What are the six permission modes in Claude Code?
- **Options**: A) read, write, execute, admin, root, sudo | B) default, acceptEdits, plan, auto, dontAsk, bypassPermissions | C) safe, normal, elevated, admin, unrestricted, god | D) view, edit, run, deploy, full, bypass
+- **Options**: A) read, write, execute, admin, root, sudo | B) manual, acceptEdits, plan, auto, dontAsk, bypassPermissions | C) safe, normal, elevated, admin, unrestricted, god | D) view, edit, run, deploy, full, bypass
 - **Correct**: B
- **Explanation**: The six modes are: default (prompts for everything), acceptEdits (auto-accepts file edits), plan (read-only analysis), auto (background classifier decides), dontAsk (auto-denies unless pre-approved), bypassPermissions (skips all checks).
+- **Explanation**: The six modes are: manual (prompts for everything — renamed from `default` in v2.1.200, which is still accepted as an alias), acceptEdits (auto-accepts file edits), plan (read-only analysis), auto (background classifier decides), dontAsk (auto-denies unless pre-approved), bypassPermissions (skips all checks).
 - **Review**: Permission Modes section

 ### Q2
@@ -844,8 +844,15 @@

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
- https://code.claude.com/docs/en/discover-plugins.md
- https://code.claude.com/docs/en/plugins-reference.md
+- https://code.claude.com/docs/en/hooks
+- https://code.claude.com/docs/en/skills
+- https://code.claude.com/docs/en/mcp
+- https://code.claude.com/docs/en/memory
+- https://code.claude.com/docs/en/sub-agents
+- https://code.claude.com/docs/en/permission-modes
+- https://code.claude.com/docs/en/checkpointing
+- https://code.claude.com/docs/en/plugins-reference
+- https://code.claude.com/docs/en/cli-reference
@@ -33,7 +33,7 @@ The output format for Step 5 ("Score and Present Results"). Fill the bracketed p
 **Pre-test score: N/10.** This gives you a baseline! Focus your study on the topics you missed. After completing the lesson, retake the quiz to measure your improvement.

 [If during]:
-**Progress check: N/10.** [If 7+: Great progress — keep going! If 4-6: Review the incorrect topics before continuing. If <4: Consider re-reading from the beginning.]
+**Progress check: N/10.** [If 7 or above: Great progress — keep going! If 5-6: Review the incorrect topics before continuing. If 0-4: Consider re-reading from the beginning.]

 [If after]:
 **Mastery check: N/10.** [If 9-10: You've mastered this lesson! Move on to the next. If 7-8: Almost there — review the missed topics and retake. If <7: Spend more time with the lesson, especially the sections marked above.]
@@ -46,3 +46,10 @@ The output format for Step 5 ("Score and Present Results"). Fill the bracketed p
 - [If developing or below]: Re-read the full lesson: [lesson link]. Focus on: [list weak categories]
 - [Offer]: "Would you like to retake this quiz, try a different lesson, or get help with a specific topic?"
 ```
+
+---
+
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
+**Sources**:
+- https://code.claude.com/docs/en/overview
@@ -42,7 +42,7 @@ graph TD

 ### Deep Assessment (~5 min)
 - 5 rounds of questions covering 10 feature areas (2 topics per round)
- Per-topic scoring (0-2 points each, 20 points total)
+- Per-topic scoring (0-2 points each, 19 points maximum — Checkpoints contributes 0-1)
 - Mastery table with strength areas, priority gaps, and review items
 - Dependency-aware learning path with phases and time estimates
 - Recommended practice projects combining gap topics
@@ -1,6 +1,6 @@
 ---
 name: self-assessment
-version: 2.3.0
+version: 2.4.0
 description: Comprehensive Claude Code self-assessment and learning path advisor. Runs a multi-category quiz covering 10 feature areas, produces a detailed skill profile with per-topic scores, identifies specific gaps, and generates a personalized learning path with prioritized next steps. Use when asked to "assess my level", "take the quiz", "find my level", "where should I start", "what should I learn next", "check my skills", "skill check", or "level up".
 ---

@@ -67,7 +67,7 @@ Options:
 1. "Created a custom slash command or skill" — Written a SKILL.md file with frontmatter, or created .claude/commands/ files
 2. "Used dynamic context in commands" — Used `$ARGUMENTS`, `$0`/`$1`, backtick `!command` syntax, or `@file` references in skill/command files
 3. "Set up project + personal memory" — Created both a project CLAUDE.md and personal ~/.claude/CLAUDE.md (or CLAUDE.local.md)
-4. "Used memory hierarchy features" — Understand the 7-level priority order, used .claude/rules/ directory, path-specific rules, or @import syntax
+4. "Used memory hierarchy features" — Understand how the 7 memory locations are concatenated into context (loaded root-down, CLAUDE.local.md appended after CLAUDE.md at each level, subdirectory files loaded on demand), used .claude/rules/ directory, path-specific rules, or @import syntax

 **Scoring for Round 1:**
 - Options 1-2 map to **Slash Commands** (0-2 points)
@@ -97,7 +97,7 @@ Options:
 1. "Connected an MCP server and used its tools" — e.g., GitHub MCP for PRs/issues, database MCP for queries, or any external data source
 2. "Used advanced MCP features" — Project-scope .mcp.json, OAuth authentication, MCP resources with @mentions, Tool Search, or `claude mcp serve`
 3. "Created or configured custom subagents" — Defined agents in .claude/agents/ with custom tools, model, or permissions
-4. "Used advanced subagent features" — Worktree isolation, persistent agent memory, background tasks with Ctrl+B, agent allowlists with `Task(agent_name)`, or agent teams
+4. "Used advanced subagent features" — Worktree isolation, persistent agent memory, background tasks with Ctrl+B, agent allowlists with `Agent(agent_type)` (the `Task(...)` form is a back-compat alias), or agent teams

 **Scoring for Round 3:**
 - Options 1-2 map to **MCP** (0-2 points)
@@ -109,9 +109,9 @@ Options:

 "Which of these have you done? Select all that apply."
 Options:
-1. "Used checkpoints for safe experimentation" — Created checkpoints, used Esc+Esc or /rewind, restored code and/or conversation, or used Summarize option
+1. "Used checkpoints for safe experimentation" — Created checkpoints, used Esc+Esc or /rewind, restored code and/or conversation, or used either Summarize option (summarize from here / summarize up to here)
 2. "Used planning mode or extended thinking" — Activated planning via /plan, Shift+Tab, or --permission-mode plan; toggled extended thinking with Alt+T/Option+T
-3. "Configured permission modes" — Used acceptEdits, plan, dontAsk, or bypassPermissions mode via CLI flags, keyboard shortcuts, or settings
+3. "Configured permission modes" — Used any of the six modes — manual (renamed from default in v2.1.200), acceptEdits, plan, auto, dontAsk, or bypassPermissions — via CLI flags, keyboard shortcuts, or settings
 4. "Used remote/desktop/web features" — Used `claude remote-control`, `claude --remote`, `/teleport`, `/desktop`, or worktrees with `claude -w`

 **Scoring for Round 4:**
@@ -125,7 +125,7 @@ Options:
 "Which of these have you done? Select all that apply."
 Options:
 1. "Installed or created a plugin" — Used a bundled plugin from marketplace, or created a .claude-plugin/ directory with plugin.json manifest
-2. "Used plugin advanced features" — Plugin hooks, plugin MCP servers, LSP configuration, plugin namespaced commands, or --plugin-dir flag for testing
+2. "Used plugin advanced features" — Plugin skills (`skills/` — the preferred form; namespaced `commands/` is legacy but still works), plugin hooks, plugin MCP servers, LSP configuration, or --plugin-dir flag for testing
 3. "Used print mode in scripts or CI/CD" — Used `claude -p` with --output-format json, --max-turns, piped input, or integrated into GitHub Actions / CI pipelines
 4. "Used advanced CLI features" — Session resumption (-c/-r), --agents flag, --json-schema for structured output, --fallback-model, --from-pr, or batch processing loops

@@ -181,7 +181,7 @@ Calculate per-topic scores from the 5 rounds. Each topic gets 0-2 points. Then p

 ### Overall Level: [Level 1 / Level 2 / Level 3]

-**Total Score: N/20 points**
+**Total Score: N/19 points**

 [One-line motivational summary]

@@ -219,7 +219,7 @@ Calculate per-topic scores from the 5 rounds. Each topic gets 0-2 points. Then p
 **Overall level calculation for Deep Assessment:**
 - 0-6 total points = Level 1: Beginner
 - 7-13 total points = Level 2: Intermediate
- 14-20 total points = Level 3: Advanced
+- 14-19 total points = Level 3: Advanced

 ---

@@ -297,8 +297,8 @@ Use these specific recommendations when a topic is a gap:
 - Done when: Claude remembers your preferences across sessions

 **Memory (score 1 — review)**:
- Focus on: 7-level hierarchy and priority order, .claude/rules/ directory with path-specific rules, `@import` syntax (max depth 5), Auto Memory MEMORY.md (200-line limit)
- Done when: You have modular rules for different directories and understand the full hierarchy
+- Focus on: the 7 memory locations and how they are concatenated into context (loaded root-down, not overridden), .claude/rules/ directory with path-specific rules, `@import` syntax (max depth 4), Auto Memory MEMORY.md (Claude loads its first 200 lines or 25 KB, whichever comes first — a load cap, not a file-size limit)
+- Done when: You have modular rules for different directories and understand how every memory file is concatenated into context

 **Skills (score 0)**:
 - Tutorial: [03-skills/](../../../03-skills/)
@@ -307,7 +307,7 @@ Use these specific recommendations when a topic is a gap:
 - Done when: A skill automatically activates based on conversation context

 **Skills (score 1 — review)**:
- Focus on: `context: fork` with `agent` field for subagent execution, `disable-model-invocation` vs `user-invocable`, 2% context budget, bundled resources (scripts/, references/, assets/)
+- Focus on: `context: fork` with `agent` field for subagent execution, `disable-model-invocation` vs `user-invocable`, the skill-listing budget (1% of the context window, fallback 8,000 characters, 250 characters per entry), bundled resources (scripts/, references/, assets/)
 - Done when: You can create a skill that runs in a subagent with forked context

 **Hooks (score 0)**:
@@ -317,38 +317,38 @@ Use these specific recommendations when a topic is a gap:
 - Done when: A hook blocks dangerous commands before execution

 **Hooks (score 1 — review)**:
- Focus on: All 25 hook events (including PostToolUseFailure, StopFailure, TaskCreated, CwdChanged, FileChanged, PostCompact, Elicitation, ElicitationResult), 4 hook types (command, http, prompt, agent), component-scoped hooks in SKILL.md frontmatter, HTTP hooks with allowedEnvVars, `CLAUDE_ENV_FILE` for SessionStart/CwdChanged/FileChanged
+- Focus on: All 31 hook events (including PostToolUseFailure, StopFailure, TaskCreated, CwdChanged, FileChanged, PostCompact, Elicitation, ElicitationResult, Setup, UserPromptExpansion, MessageDisplay), 5 hook types (command, http, mcp_tool, prompt, agent — agent hooks are experimental and may change), component-scoped hooks in SKILL.md frontmatter, HTTP hooks with allowedEnvVars, `CLAUDE_ENV_FILE` for SessionStart/CwdChanged/FileChanged
 - Done when: You can create a prompt-based Stop hook and a component-scoped hook in a skill

 **MCP (score 0)**:
 - Tutorial: [05-mcp/](../../../05-mcp/)
- Focus on: `claude mcp add` command, transport types (HTTP recommended), GitHub MCP setup, environment variable expansion
+- Focus on: `claude mcp add` command, transport types (`http` recommended, `stdio`, `ws` for push-style servers, and the deprecated `sse` — note `--transport` does not accept `ws`, so add WebSocket servers with `claude mcp add-json`), GitHub MCP setup, environment variable expansion
 - Key exercise: Add GitHub MCP server and query PRs
 - Done when: You can query live data from an external service via MCP

 **MCP (score 1 — review)**:
- Focus on: Project-scope .mcp.json (requires team approval), OAuth 2.0 auth, MCP resources with `@server:resource` mentions, Tool Search (ENABLE_TOOL_SEARCH), `claude mcp serve`, output limits (10k/25k/50k)
+- Focus on: Project-scope .mcp.json (requires team approval), OAuth 2.0 auth, MCP resources with `@server:resource` mentions, Tool Search (ENABLE_TOOL_SEARCH), `claude mcp serve`, output limits (10,000 tokens warning; 25,000 tokens default max via `MAX_MCP_OUTPUT_TOKENS`; 50,000 characters disk-persistence threshold)
 - Done when: You have a project .mcp.json and understand Tool Search auto mode

 **Subagents (score 0)**:
 - Tutorial: [04-subagents/](../../../04-subagents/)
- Focus on: Agent file format (.claude/agents/*.md), built-in agents (general-purpose, Plan, Explore), tools/model/permissionMode config
+- Focus on: Agent file format (.claude/agents/*.md), built-in agents (Explore, Plan, general-purpose, claude, statusline-setup, claude-code-guide), tools/model/permissionMode config, spawn limits (depth default 3 since v2.1.219, concurrency default 20, and the 200-per-session spawn cap removed in v2.1.224)
 - Key exercise: Create a code-reviewer subagent and test delegation
 - Done when: Claude delegates code review to your custom agent

 **Subagents (score 1 — review)**:
- Focus on: Worktree isolation (`isolation: worktree`), persistent agent memory (`memory` field with scopes), background agents (Ctrl+B/Ctrl+F), agent allowlists with `Task(agent_name)`, agent teams (`--teammate-mode`)
+- Focus on: Worktree isolation (`isolation: worktree`), persistent agent memory (`memory` field with scopes), background agents (Ctrl+B/Ctrl+F), agent allowlists with `Agent(agent_type)` (`Task(...)` remains a back-compat alias), agent teams (`--teammate-mode`)
 - Done when: You have a subagent with persistent memory running in worktree isolation

 **Checkpoints (score 0)**:
 - Tutorial: [08-checkpoints/](../../../08-checkpoints/)
- Focus on: Esc+Esc and /rewind access, 5 rewind options (restore code+conversation, restore conversation, restore code, summarize, cancel), limitations (bash filesystem ops not tracked)
+- Focus on: Esc+Esc and /rewind access, 6 rewind options (restore code and conversation, restore conversation, restore code, summarize from here, summarize up to here, never mind), limitations (bash filesystem ops not tracked)
 - Key exercise: Make experimental changes, then rewind to restore
 - Done when: You can confidently experiment knowing you can rewind

 **Advanced Features (score 0)**:
 - Tutorial: [09-advanced-features/](../../../09-advanced-features/)
- Focus on: Planning mode (/plan or Shift+Tab), permission modes (5 types), extended thinking (Alt+T toggle)
+- Focus on: Planning mode (/plan or Shift+Tab), permission modes (6 types: manual — renamed from default in v2.1.200 — acceptEdits, plan, auto, dontAsk, bypassPermissions), extended thinking (Alt+T toggle)
 - Key exercise: Use planning mode to design a feature, then implement it
 - Done when: You can switch between planning and implementation modes fluently

@@ -358,7 +358,7 @@ Use these specific recommendations when a topic is a gap:

 **Plugins (score 0)**:
 - Tutorial: [07-plugins/](../../../07-plugins/)
- Focus on: Plugin structure (.claude-plugin/plugin.json), what plugins bundle (commands, agents, MCP, hooks, settings), installation from marketplace
+- Focus on: Plugin structure (.claude-plugin/plugin.json), what plugins bundle (skills, agents, MCP, hooks, settings — plus the legacy `commands/` directory, which still works but `skills/` is preferred for new plugins), installation from marketplace
 - Key exercise: Install a plugin and explore its components
 - Done when: You understand when to use a plugin vs standalone components

@@ -11,7 +11,7 @@ Options:
 1. "Created a custom slash command or skill" — Written a SKILL.md file with frontmatter, or created .claude/commands/ files
 2. "Used dynamic context in commands" — Used `$ARGUMENTS`, `$0`/`$1`, backtick `!command` syntax, or `@file` references in skill/command files
 3. "Set up project + personal memory" — Created both a project CLAUDE.md and personal ~/.claude/CLAUDE.md (or CLAUDE.local.md)
-4. "Used memory hierarchy features" — Understand the 7-level priority order, used .claude/rules/ directory, path-specific rules, or @import syntax
+4. "Used memory hierarchy features" — Understand how the 7 memory locations are concatenated into context (loaded root-down, CLAUDE.local.md appended after CLAUDE.md at each level, subdirectory files loaded on demand), used .claude/rules/ directory, path-specific rules, or @import syntax

 **Scoring:** Options 1-2 → **Slash Commands** (0-2); Options 3-4 → **Memory** (0-2)

@@ -37,7 +37,7 @@ Options:
 1. "Connected an MCP server and used its tools" — e.g., GitHub MCP for PRs/issues, database MCP for queries, or any external data source
 2. "Used advanced MCP features" — Project-scope .mcp.json, OAuth authentication, MCP resources with @mentions, Tool Search, or `claude mcp serve`
 3. "Created or configured custom subagents" — Defined agents in .claude/agents/ with custom tools, model, or permissions
-4. "Used advanced subagent features" — Worktree isolation, persistent agent memory, background tasks with Ctrl+B, agent allowlists with `Task(agent_name)`, or agent teams
+4. "Used advanced subagent features" — Worktree isolation, persistent agent memory, background tasks with Ctrl+B, agent allowlists with `Agent(agent_type)` (the `Task(...)` form is a back-compat alias), or agent teams

 **Scoring:** Options 1-2 → **MCP** (0-2); Options 3-4 → **Subagents** (0-2)

@@ -47,9 +47,9 @@ Options:

 "Which of these have you done? Select all that apply."
 Options:
-1. "Used checkpoints for safe experimentation" — Created checkpoints, used Esc+Esc or /rewind, restored code and/or conversation, or used Summarize option
+1. "Used checkpoints for safe experimentation" — Created checkpoints, used Esc+Esc or /rewind, restored code and/or conversation, or used either Summarize option (summarize from here / summarize up to here)
 2. "Used planning mode or extended thinking" — Activated planning via /plan, Shift+Tab, or --permission-mode plan; toggled extended thinking with Alt+T/Option+T
-3. "Configured permission modes" — Used acceptEdits, plan, dontAsk, or bypassPermissions mode via CLI flags, keyboard shortcuts, or settings
+3. "Configured permission modes" — Used any of the six modes — manual (renamed from default in v2.1.200), acceptEdits, plan, auto, dontAsk, or bypassPermissions — via CLI flags, keyboard shortcuts, or settings
 4. "Used remote/desktop/web features" — Used `claude remote-control`, `claude --remote`, `/teleport`, `/desktop`, or worktrees with `claude -w`

 **Scoring:** Option 1 → **Checkpoints** (0-1); Options 2-4 → **Advanced Features** (0-3, cap at 2)
@@ -61,8 +61,19 @@ Options:
 "Which of these have you done? Select all that apply."
 Options:
 1. "Installed or created a plugin" — Used a bundled plugin from marketplace, or created a .claude-plugin/ directory with plugin.json manifest
-2. "Used plugin advanced features" — Plugin hooks, plugin MCP servers, LSP configuration, plugin namespaced commands, or --plugin-dir flag for testing
+2. "Used plugin advanced features" — Plugin skills (`skills/` — the preferred form; namespaced `commands/` is legacy but still works), plugin hooks, plugin MCP servers, LSP configuration, or --plugin-dir flag for testing
 3. "Used print mode in scripts or CI/CD" — Used `claude -p` with --output-format json, --max-turns, piped input, or integrated into GitHub Actions / CI pipelines
 4. "Used advanced CLI features" — Session resumption (-c/-r), --agents flag, --json-schema for structured output, --fallback-model, --from-pr, or batch processing loops

 **Scoring:** Options 1-2 → **Plugins** (0-2); Options 3-4 → **CLI** (0-2)
+
+---
+
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
+**Sources**:
+- https://code.claude.com/docs/en/memory
+- https://code.claude.com/docs/en/sub-agents
+- https://code.claude.com/docs/en/permission-modes
+- https://code.claude.com/docs/en/checkpointing
+- https://code.claude.com/docs/en/plugins-reference
@@ -42,7 +42,7 @@ You checked **N/8** items.

 ### Overall Level: [Level 1 / Level 2 / Level 3]

-**Total Score: N/20 points**
+**Total Score: N/19 points**

 [One-line motivational summary]

@@ -113,7 +113,7 @@ Based on your gaps, try these real-world exercises to solidify your learning:

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
- internal consistency fix — no external source
+- https://code.claude.com/docs/en/overview
@@ -19,8 +19,8 @@ Use these specific recommendations when a topic is a gap. Paths are relative to
 - Done when: Claude remembers your preferences across sessions

 **Memory (score 1 — review)**:
- Focus on: 7-level hierarchy and priority order, .claude/rules/ directory with path-specific rules, `@import` syntax (max depth 5), Auto Memory MEMORY.md (200-line limit)
- Done when: You have modular rules for different directories and understand the full hierarchy
+- Focus on: the 7 memory locations and how they are concatenated into context (loaded root-down, not overridden), .claude/rules/ directory with path-specific rules, `@import` syntax (max depth 4), Auto Memory MEMORY.md (Claude loads its first 200 lines or 25 KB, whichever comes first — a load cap, not a file-size limit)
+- Done when: You have modular rules for different directories and understand how every memory file is concatenated into context

 **Skills (score 0)**:
 - Tutorial: [03-skills/](../../../../03-skills/)
@@ -29,7 +29,7 @@ Use these specific recommendations when a topic is a gap. Paths are relative to
 - Done when: A skill automatically activates based on conversation context

 **Skills (score 1 — review)**:
- Focus on: `context: fork` with `agent` field for subagent execution, `disable-model-invocation` vs `user-invocable`, 2% context budget, bundled resources (scripts/, references/, assets/)
+- Focus on: `context: fork` with `agent` field for subagent execution, `disable-model-invocation` vs `user-invocable`, the skill-listing budget (1% of the context window, fallback 8,000 characters, 250 characters per entry), bundled resources (scripts/, references/, assets/)
 - Done when: You can create a skill that runs in a subagent with forked context

 **Hooks (score 0)**:
@@ -39,38 +39,38 @@ Use these specific recommendations when a topic is a gap. Paths are relative to
 - Done when: A hook blocks dangerous commands before execution

 **Hooks (score 1 — review)**:
- Focus on: All 25 hook events (including PostToolUseFailure, StopFailure, TaskCreated, CwdChanged, FileChanged, PostCompact, Elicitation, ElicitationResult), 4 hook types (command, http, prompt, agent), component-scoped hooks in SKILL.md frontmatter, HTTP hooks with allowedEnvVars, `CLAUDE_ENV_FILE` for SessionStart/CwdChanged/FileChanged
+- Focus on: All 31 hook events (including PostToolUseFailure, StopFailure, TaskCreated, CwdChanged, FileChanged, PostCompact, Elicitation, ElicitationResult, Setup, UserPromptExpansion, MessageDisplay), 5 hook types (command, http, mcp_tool, prompt, agent — agent hooks are experimental and may change), component-scoped hooks in SKILL.md frontmatter, HTTP hooks with allowedEnvVars, `CLAUDE_ENV_FILE` for SessionStart/CwdChanged/FileChanged
 - Done when: You can create a prompt-based Stop hook and a component-scoped hook in a skill

 **MCP (score 0)**:
 - Tutorial: [05-mcp/](../../../../05-mcp/)
- Focus on: `claude mcp add` command, transport types (HTTP recommended), GitHub MCP setup, environment variable expansion
+- Focus on: `claude mcp add` command, transport types (`http` recommended, `stdio`, `ws` for push-style servers, and the deprecated `sse` — note `--transport` does not accept `ws`, so add WebSocket servers with `claude mcp add-json`), GitHub MCP setup, environment variable expansion
 - Key exercise: Add GitHub MCP server and query PRs
 - Done when: You can query live data from an external service via MCP

 **MCP (score 1 — review)**:
- Focus on: Project-scope .mcp.json (requires team approval), OAuth 2.0 auth, MCP resources with `@server:resource` mentions, Tool Search (ENABLE_TOOL_SEARCH), `claude mcp serve`, output limits (10k/25k/50k)
+- Focus on: Project-scope .mcp.json (requires team approval), OAuth 2.0 auth, MCP resources with `@server:resource` mentions, Tool Search (ENABLE_TOOL_SEARCH), `claude mcp serve`, output limits (10,000 tokens warning; 25,000 tokens default max via `MAX_MCP_OUTPUT_TOKENS`; 50,000 characters disk-persistence threshold)
 - Done when: You have a project .mcp.json and understand Tool Search auto mode

 **Subagents (score 0)**:
 - Tutorial: [04-subagents/](../../../../04-subagents/)
- Focus on: Agent file format (.claude/agents/*.md), built-in agents (general-purpose, Plan, Explore), tools/model/permissionMode config
+- Focus on: Agent file format (.claude/agents/*.md), built-in agents (Explore, Plan, general-purpose, claude, statusline-setup, claude-code-guide), tools/model/permissionMode config, spawn limits (depth default 3 since v2.1.219, concurrency default 20, and the 200-per-session spawn cap removed in v2.1.224)
 - Key exercise: Create a code-reviewer subagent and test delegation
 - Done when: Claude delegates code review to your custom agent

 **Subagents (score 1 — review)**:
- Focus on: Worktree isolation (`isolation: worktree`), persistent agent memory (`memory` field with scopes), background agents (Ctrl+B/Ctrl+F), agent allowlists with `Task(agent_name)`, agent teams (`--teammate-mode`)
+- Focus on: Worktree isolation (`isolation: worktree`), persistent agent memory (`memory` field with scopes), background agents (Ctrl+B/Ctrl+F), agent allowlists with `Agent(agent_type)` (`Task(...)` remains a back-compat alias), agent teams (`--teammate-mode`)
 - Done when: You have a subagent with persistent memory running in worktree isolation

 **Checkpoints (score 0)**:
 - Tutorial: [08-checkpoints/](../../../../08-checkpoints/)
- Focus on: Esc+Esc and /rewind access, 5 rewind options (restore code+conversation, restore conversation, restore code, summarize, cancel), limitations (bash filesystem ops not tracked)
+- Focus on: Esc+Esc and /rewind access, 6 rewind options (restore code and conversation, restore conversation, restore code, summarize from here, summarize up to here, never mind), limitations (bash filesystem ops not tracked)
 - Key exercise: Make experimental changes, then rewind to restore
 - Done when: You can confidently experiment knowing you can rewind

 **Advanced Features (score 0)**:
 - Tutorial: [09-advanced-features/](../../../../09-advanced-features/)
- Focus on: Planning mode (/plan or Shift+Tab), permission modes (5 types), extended thinking (Alt+T toggle)
+- Focus on: Planning mode (/plan or Shift+Tab), permission modes (6 types: manual — renamed from default in v2.1.200 — acceptEdits, plan, auto, dontAsk, bypassPermissions), extended thinking (Alt+T toggle)
 - Key exercise: Use planning mode to design a feature, then implement it
 - Done when: You can switch between planning and implementation modes fluently

@@ -80,7 +80,7 @@ Use these specific recommendations when a topic is a gap. Paths are relative to

 **Plugins (score 0)**:
 - Tutorial: [07-plugins/](../../../../07-plugins/)
- Focus on: Plugin structure (.claude-plugin/plugin.json), what plugins bundle (commands, agents, MCP, hooks, settings), installation from marketplace
+- Focus on: Plugin structure (.claude-plugin/plugin.json), what plugins bundle (skills, agents, MCP, hooks, settings — plus the legacy `commands/` directory, which still works but `skills/` is preferred for new plugins), installation from marketplace
 - Key exercise: Install a plugin and explore its components
 - Done when: You understand when to use a plugin vs standalone components

@@ -97,3 +97,17 @@ Use these specific recommendations when a topic is a gap. Paths are relative to
 **CLI (score 1 — review)**:
 - Focus on: --agents flag with JSON config, --json-schema for structured output, --fallback-model, --from-pr, --strict-mcp-config, batch processing with for loops, `claude mcp serve`
 - Done when: You have a CI/CD script that uses Claude with structured JSON output
+
+---
+
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
+**Sources**:
+- https://code.claude.com/docs/en/hooks
+- https://code.claude.com/docs/en/memory
+- https://code.claude.com/docs/en/skills
+- https://code.claude.com/docs/en/mcp
+- https://code.claude.com/docs/en/sub-agents
+- https://code.claude.com/docs/en/checkpointing
+- https://code.claude.com/docs/en/permission-modes
+- https://code.claude.com/docs/en/plugins-reference
@@ -20,6 +20,8 @@ Slash commands are shortcuts that control Claude's behavior during an interactiv

 Built-in commands are shortcuts for common actions. There are **60+ built-in commands** and **10 bundled skills** available. Type `/` in Claude Code to see the full list, or type `/` followed by any letters to filter.

+> **Note**: Since v2.1.236, pressing Enter on a mistyped slash command — or on a command that is not available in the current session — reports an error instead of silently running the closest fuzzy match. Unambiguous prefixes and defined aliases still run as before.
+
 | Command | Purpose |
 |---------|---------|
 | `/add-dir <path>` | Add working directory |
@@ -96,7 +98,6 @@ Built-in commands are shortcuts for common actions. There are **60+ built-in com
 | `/theme` | Open theme picker / manage custom themes (v2.1.118). Define custom themes via JSON in `~/.claude/themes/<name>.json` |
 | `/tui` | Toggle fullscreen TUI (text user interface) mode with flicker-free rendering (added v2.1.110) |
 | `/ultrareview` | Comprehensive cloud-based multi-agent code review (added v2.1.111). The preferred invocation is now `/code-review ultra`; `/ultrareview` remains as an alias. Includes 3 free runs on Pro and Max, then requires usage credits |
-| `/undo` | Alias for `/rewind` (added v2.1.108) |
 | `/upgrade` | Open upgrade page for higher plan tier |
 | `/usage` | Canonical usage dashboard (v2.1.118) — combines plan usage limits, rate limits, cost, and daily session stats. `/cost` and `/stats` are typing-shortcut aliases that open specific tabs |
 | `/voice` | Toggle push-to-talk voice dictation |
@@ -114,6 +115,7 @@ These skills ship with Claude Code and are invoked like slash commands:
 | `/claude-api` | Load Claude API reference for project language |
 | `/dataviz` | Chart and dashboard design guidance with a runnable color-palette validator (v2.1.198) |
 | `/debug [description]` | Enable debug logging |
+| `/design [description]` | Create a **design canvas** — a multi-artboard visual design (UI mockups, screen flows, landing pages, posters) published as an artifact and refined visually rather than in code. Use it instead of hand-writing HTML when the layout is what you want to iterate on. Research preview; requires v2.1.233+ and a Pro, Max, Team, or Enterprise plan |
 | `/loop [interval] <prompt>` | Run prompt repeatedly on interval |
 | `/code-review [low\|medium\|high\|xhigh\|max\|ultra] [--fix] [--comment] [pr#\|branch\|path]` | Review the current diff — or a PR number, branch, or path you pass — for correctness bugs. Pass `--fix` to apply findings, `--comment` to post them as inline GitHub PR comments, or `ultra` to run a deep cloud review; with `ultra` on a `github.com` PR target, `--post` preselects posting the findings to the PR. With no effort level given, the review reuses the last level you typed (v2.1.223). Originally absorbed `/simplify` in v2.1.146, but `/simplify` returned as a distinct command in v2.1.154 |
 | `/simplify` | Run a cleanup-only review (reuse / simplification / efficiency / altitude) and apply the fixes; does **not** hunt for bugs — use `/code-review` for that. Briefly an alias of `/code-review --fix` (v2.1.152), it became cleanup-only in v2.1.154 |
@@ -125,12 +127,13 @@ These skills ship with Claude Code and are invoked like slash commands:
 | `/output-style` | Removed in v2.1.91 (deprecated v2.1.73) — use `/config` → Output style, or the `outputStyle` setting |
 | `/pr-comments` | Removed in v2.1.91 — ask Claude directly to view PR comments |
 | `/vim` | Removed in v2.1.92 — use /config → Editor mode |
+| `/undo` | No longer listed in the official commands reference as of v2.1.245 (it was added as an alias for `/rewind` in v2.1.108) — use `/rewind` or press `Esc` twice |

 ### Recent Changes

 - `/fork` and `/subtask` swapped roles in **v2.1.212**. `/fork` now copies the conversation into a new independent background session; the forked-subagent behavior it used to have moved to the new `/subtask` command. History: `/fork` was an alias for `/branch` from v2.1.77 to v2.1.161; from v2.1.161 to v2.1.211 it started a forked subagent (what `/subtask` does now). When agent view is turned off, `/subtask` is unavailable and `/fork` keeps the forked-subagent behavior
 - `/resume` (no arguments) opens a picker of past sessions — including ones removed from the visible list — and resumes the chosen one as a background session (v2.1.212)
- `/output-style` deprecated (v2.1.73) and removed (v2.1.91) — output styles are still available via `/config` → Output style or the `outputStyle` setting; the built-ins are Default, Proactive, Explanatory, and Learning
+- `/output-style` deprecated (v2.1.73) and removed (v2.1.91) — output styles are still available via `/config` → Output style or the `outputStyle` setting; the built-ins are Default, Proactive, Explanatory, Learning, and Concise (added in v2.1.237)
 - `/review` became a full alias of `/code-review` — same targets, effort levels, and flags (v2.1.223). History: it first moved onto the `/code-review medium` engine in v2.1.186 while remaining PR-only
 - `/effort` command added; `max` level available on Opus 4.6+ (originally Opus 4.6-only)
 - `/voice` command added for push-to-talk voice dictation
@@ -148,7 +151,7 @@ These skills ship with Claude Code and are invoked like slash commands:
 - `/tui` command added for flicker-free fullscreen TUI rendering (v2.1.110)
 - `/focus` command added for focus view toggle; `Ctrl+O` now only toggles verbose transcript (v2.1.110)
 - `/recap` command added to manually trigger session context recap (v2.1.108)
- `/undo` added as alias for `/rewind` (v2.1.108)
+- `/undo` added as alias for `/rewind` (v2.1.108); it no longer appears in the official commands reference as of v2.1.245 — use `/rewind` or `Esc Esc`
 - `/proactive` added as alias for `/loop` (v2.1.105)
 - `/effort` gained interactive arrow-key slider and new `xhigh` level between `high` and `max`; default effort raised to `xhigh` for Opus 4.7 plans (v2.1.111). On Opus 4.8 the default is `high` (v2.1.154); Opus 5 also defaults to `high` (v2.1.219)
 - `/ultrareview` added for comprehensive cloud-based multi-agent code review (v2.1.111)
@@ -645,13 +648,14 @@ If both exist with the same name, the **skill takes precedence**. Remove one or

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/skills
 - https://code.claude.com/docs/en/interactive-mode
 - https://code.claude.com/docs/en/changelog
 - https://code.claude.com/docs/en/commands
+- https://code.claude.com/docs/en/whats-new/2026-w34
 - https://code.claude.com/docs/en/model-config
 - https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.139
@@ -683,7 +683,9 @@ Claude will load CLAUDE.md from the specified additional directory alongside the
 ````markdown
 # API Module Standards

-This file overrides root CLAUDE.md for everything in /src/api/
+This file supplements root CLAUDE.md for everything in /src/api/. Memory files are
+concatenated, not overridden — the root CLAUDE.md still applies, and Claude Code
+loads this file on demand when it reads files in this subtree.

 ## API-Specific Standards

@@ -1113,7 +1115,9 @@ If you prefer manual setup:
   cat > /path/to/directory/CLAUDE.md << 'EOF'
   # [Directory Name] Standards

-   This file overrides root CLAUDE.md for this directory.
+   This file supplements root CLAUDE.md for this directory. Memory files are
+   concatenated, not overridden — Claude Code loads this file on demand when it
+   reads files in this directory.

   ## [Specific Standards]
   EOF
@@ -1189,8 +1193,8 @@ Auto Memory is a separate mechanism (`~/.claude/projects/<project>/memory/`), no

 ---

-**Last Updated**: August 4, 2026
-**Claude Code Version**: 2.1.220
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/memory
 **Compatible Models**: Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -93,7 +93,7 @@ sequenceDiagram
 | **Project** | `.claude/skills/<skill-name>/SKILL.md` | Team | Yes (via git) | Team standards |
 | **Plugin** | `<plugin>/skills/<skill-name>/SKILL.md` | Where enabled | Depends | Bundled with plugins |

-When skills share the same name across levels, higher-priority locations win: **enterprise > project > personal**. Project skills override personal ones by default; the `skillOverrides` setting (v2.1.129+) tunes that behavior — see [Controlling Skill Override Behavior](#controlling-skill-override-behavior-skilloverrides). Plugin skills use a `plugin-name:skill-name` namespace, so they cannot conflict.
+When skills share the same name across levels, higher-priority locations win: **enterprise > personal > project**. Personal skills override project ones by default; the `skillOverrides` setting (v2.1.129+) tunes that behavior — see [Controlling Skill Override Behavior](#controlling-skill-override-behavior-skilloverrides). Plugin skills use a `plugin-name:skill-name` namespace, so they cannot conflict.

 > **Subagent skill discovery (v2.1.133+)**: Subagents now discover project, user, and plugin skills via the Skill tool the same way the main session does. Earlier versions limited subagents to their own embedded set, which meant skill+subagent workflows quietly degraded; from v2.1.133 the same skill catalog is visible to both.

@@ -141,7 +141,7 @@ Show concrete examples of using this Skill.
 ### Required Fields

 - **name**: lowercase letters, numbers, hyphens only (max 64 characters). Cannot contain "anthropic" or "claude".
- **description**: what the Skill does AND when to use it (max 1024 characters). This is critical for Claude to know when to activate the skill.
+- **description**: what the Skill does AND when to use it. The combined `description` + `when_to_use` text is truncated at **1,536 characters** in the skill listing (configurable via `skillListingMaxDescChars`). This is critical for Claude to know when to activate the skill.

 ### Optional Frontmatter Fields

@@ -173,7 +173,8 @@ paths: "src/api/**/*.ts"               # Glob patterns limiting when skill activ
 | Field | Description |
 |-------|-------------|
 | `name` | Lowercase letters, numbers, hyphens only (max 64 chars). Cannot contain "anthropic" or "claude". |
-| `description` | What the Skill does AND when to use it (max 1024 chars). Critical for auto-invocation matching. |
+| `description` | What the Skill does AND when to use it. The combined `description` + `when_to_use` text is truncated at 1,536 chars in the skill listing (configurable via `skillListingMaxDescChars`). Critical for auto-invocation matching. |
+| `when_to_use` | Additional context for when Claude should invoke the skill. Appended to `description` in the skill listing and counts toward the 1,536-character cap. |
 | `argument-hint` | Hint shown in the `/` autocomplete menu (e.g., `"[filename] [format]"`). |
 | `disable-model-invocation` | `true` = only the user can invoke via `/name`. Claude will never auto-invoke. |
 | `user-invocable` | `false` = hidden from the `/` menu. Only Claude can invoke it automatically. |
@@ -187,6 +188,12 @@ paths: "src/api/**/*.ts"               # Glob patterns limiting when skill activ
 | `shell` | Shell used for `` !`command` `` substitutions and scripts: `bash` (default) or `powershell`. |
 | `hooks` | Hooks scoped to this skill's lifecycle (same format as global hooks). |
 | `paths` | Glob patterns that limit when the skill is auto-activated. Comma-separated string or YAML list. Same format as path-specific rules. |
+| `arguments` | Declares the arguments the skill accepts, for autocomplete and argument substitution. |
+| `metadata` | Free-form key/value map for your own bookkeeping (e.g. `version`, `author`). Claude Code passes it through. |
+| `license` | License identifier for the skill (e.g. `MIT`). |
+| `compatibility` | Free-text compatibility statement, up to 500 characters. Claude Code accepts it but does not act on it. |
+
+> **Note**: Only `name`, `description`, `license`, `compatibility`, `metadata`, and `allowed-tools` are valid for skills uploaded to claude.ai or created through the Skills API. The other fields in this table are Claude Code-specific.

 Since v2.1.218, boolean frontmatter fields also accept `yes`/`no`, `on`/`off`, and `1`/`0` (case-insensitive) in addition to `true`/`false`.

@@ -636,7 +643,7 @@ Can you help me review this code for security issues?

 ### Updating a Skill

-Edit the `SKILL.md` file directly. Changes take effect on next Claude Code startup.
+Edit the `SKILL.md` file directly, then run `/reload-skills` (v2.1.152+) to re-scan the skill directories. Restarting Claude Code also works, but is not required — skills in `--add-dir` directories are picked up live, and a `SessionStart` hook returning `reloadSkills: true` triggers the same re-scan.

 ```bash
 # Personal Skill
@@ -888,8 +895,8 @@ Once you start building skills seriously, two things become essential: a library

 ---

-**Last Updated**: August 4, 2026
-**Claude Code Version**: 2.1.220
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/skills
 - https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
@@ -130,7 +130,7 @@ to solving problems.
 | `tools` | No | Comma-separated list of specific tools. Omit to inherit all tools. Supports `Agent(agent_name)` syntax to restrict spawnable subagents |
 | `disallowedTools` | No | Comma-separated list of tools the subagent must not use |
 | `model` | No | Model to use: `sonnet`, `opus`, `haiku`, full model ID, or `inherit`. Defaults to configured subagent model |
-| `permissionMode` | No | `default`, `acceptEdits`, `dontAsk`, `bypassPermissions`, `plan`. As of v2.1.212, the Task tool's `mode` invocation parameter is deprecated and ignored — subagents inherit the parent session's permission mode by default unless overridden here |
+| `permissionMode` | No | `manual` (renamed from `default` in v2.1.200 — `default` is still accepted as the older name), `acceptEdits`, `dontAsk`, `bypassPermissions`, `plan`, `auto`. As of v2.1.212, the Task tool's `mode` invocation parameter is deprecated and ignored — subagents inherit the parent session's permission mode by default unless overridden here |
 | `maxTurns` | No | Maximum number of agentic turns the subagent can take |
 | `skills` | No | Comma-separated list of skills to preload. Injects full skill content into the subagent's context at startup. **v2.1.133+:** subagents also discover project, user, and plugin skills via the Skill tool — same catalog as the main session, no longer limited to their own embedded set. |
 | `mcpServers` | No | MCP servers to make available to the subagent |
@@ -238,6 +238,8 @@ claude --agents '{
 }
 ```

+> **Note**: Since v2.1.243, `--agents` no longer silently ignores invalid JSON or invalid agent definitions — Claude Code exits with a clear error, matching the behavior of `--mcp-config`.
+
 **Priority of Agent Definitions:**

 Agent definitions are loaded with this priority order (first match wins):
@@ -259,9 +261,9 @@ Claude Code includes several built-in subagents that are always available:
 | **general-purpose** | Inherits | Complex, multi-step tasks |
 | **Plan** | Inherits | Research for plan mode |
 | **Explore** | Inherits (capped at Opus) | Read-only codebase exploration (quick/medium/very thorough) |
-| **Bash** | Inherits | Terminal commands in separate context |
-| **statusline-setup** | Sonnet | Configure status line |
-| **Claude Code Guide** | Haiku | Answer Claude Code feature questions |
+| **claude** | Inherits | Catch-all for tasks that don't fit a more specialized agent; has every tool available to subagents. Also the default agent for a dispatched background session |
+| **statusline-setup** | Sonnet | Runs when you use `/statusline` to configure your status line |
+| **claude-code-guide** | Haiku | Answers questions about Claude Code features |

 ### General-Purpose Subagent

@@ -299,15 +301,15 @@ Claude Code includes several built-in subagents that are always available:
 - **"medium"** - Moderate exploration, balanced speed and thoroughness, default approach
 - **"very thorough"** - Comprehensive analysis across multiple locations and naming conventions, may take longer

-### Bash Subagent
+### Claude Subagent

 | Property | Value |
 |----------|-------|
 | **Model** | Inherits from parent |
-| **Tools** | Bash |
-| **Purpose** | Execute terminal commands in a separate context window |
+| **Tools** | Every tool available to subagents |
+| **Purpose** | Catch-all agent for tasks that don't fit a more specialized agent |

-**When used**: When running shell commands that benefit from isolated context.
+**When used**: When a task doesn't match a more specialized built-in agent. It is also the default agent for a dispatched background session; which permission mode it starts in depends on how that session was started.

 ### Statusline Setup Subagent

@@ -319,7 +321,7 @@ Claude Code includes several built-in subagents that are always available:

 **When used**: When setting up or customizing the status line.

-### Claude Code Guide Subagent
+### Claude Code Guide Subagent (`claude-code-guide`)

 | Property | Value |
 |----------|-------|
@@ -1314,8 +1316,8 @@ See the OpenTelemetry section in [Advanced Features → Telemetry](../09-advance

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/sub-agents
 - https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
@@ -128,6 +128,28 @@ Server-Sent Events transport is deprecated in favor of `http` but still supporte
 claude mcp add --transport sse legacy-server https://example.com/sse
 ```

+### WebSocket Transport (`ws`)
+
+WebSocket servers hold a persistent bidirectional connection, which suits remote MCP servers that push events to Claude unprompted. Use HTTP instead when your server only responds to requests, since HTTP supports OAuth and the `claude mcp add --transport` flag, while WebSocket supports neither.
+
+Because `--transport` does not accept `ws`, configure it in `.mcp.json` or through `claude mcp add-json`:
+
+```json
+{
+  "type": "ws",
+  "url": "wss://mcp.example.com/socket",
+  "headers": {
+    "Authorization": "Bearer YOUR_TOKEN"
+  }
+}
+```
+
+The `type: "ws"` entry accepts the same `url`, `headers`, `headersHelper`, `timeout`, and `alwaysLoad` fields as `http`. Authentication is **header-only** — there is no OAuth flow for WebSocket servers.
+
+> **Note**: WebSocket servers don't appear in `claude mcp list` output. Use `claude mcp get <name>` or the `/mcp` panel to check them.
+
+Like HTTP and SSE, WebSocket connections use a 5-minute idle window; stdio and WebSocket have no per-request timer. A `url` entry with no `type` is an error naming `"http"`, `"sse"`, and `"ws"` as the valid values.
+
 ### Session Working Directories (roots/list)

 MCP servers can discover the session's working directories: the launch directory plus all `--add-dir`/`additionalDirectories` entries are returned via the MCP `roots/list` request, and a `notifications/roots/list_changed` notification is sent whenever the set changes (v2.1.203). The idle timeout now also applies to stdio servers (30 minutes), with a per-server `timeout` acting as an idle floor (v2.1.203).
@@ -190,6 +212,8 @@ The URL must use `https://`. This option requires Claude Code v2.1.64 or later.
 - **Startup auth notice (v2.1.193+)**: At startup, Claude Code surfaces a notice listing any MCP servers that still need authentication, so a server that needs a login isn't left silently non-working.
 - **`headersHelper` auto-refresh (v2.1.193+)**: If you supply custom auth via a `headersHelper`, the helper is re-invoked automatically when a server returns HTTP 401 or 403. Credentials refresh on the fly without a manual reconnect. See [Use dynamic headers for custom authentication](https://code.claude.com/docs/en/mcp).

+> **Warning** (v2.1.238): A `headersHelper` in a project `.mcp.json`, and inline MCP servers in project or `--add-dir` agent files, now require that folder's trust dialog to have been accepted — including under `claude -p`. Such helpers also run **without inherited credential environment variables**; user-, managed-, and claude.ai-scope helpers run from the Claude config directory instead. A project config that relied on inherited credentials or on running untrusted will stop working until you accept the trust dialog and supply credentials another way.
+
 ### Claude.ai MCP Connectors

 MCP servers configured in your Claude.ai account are automatically available in Claude Code. This means any MCP connections you set up through the Claude.ai web interface will be accessible without additional configuration.
@@ -396,6 +420,8 @@ claude mcp add-json events-server '{"type":"stdio","command":"npx","args":["@mod

 > **Note**: In JSON configs — `.mcp.json`, `~/.claude.json`, or `claude mcp add-json` — the `type` field accepts `streamable-http` as an alias for `http`. The MCP specification uses the name `streamable-http` for this transport, so configurations copied from a server's own documentation work unmodified.

+Since v2.1.238, `claude mcp list` and `claude mcp get` show disabled servers as `⊘ Disabled` without connecting to them for a health check.
+
 `claude mcp login <name>` / `claude mcp logout <name>` are the non-interactive equivalent of the OAuth flow in the `/mcp` menu — authenticate or sign out without opening it. Add `--no-browser` to `login` to complete OAuth over SSH or in a headless session (it redirects the flow through stdin).

 ## Available MCP Servers Table
@@ -1252,8 +1278,8 @@ export GITHUB_TOKEN="your_token"

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/mcp
 - https://code.claude.com/docs/en/changelog
@@ -57,8 +57,14 @@ Hooks are configured in settings files with a specific structure:
 | `hooks` | Array of hook definitions | `[{ "type": "command", ... }]` |
 | `type` | Hook type: `"command"` (bash), `"prompt"` (LLM), `"http"` (webhook), `"mcp_tool"` (MCP tool invocation, v2.1.118+), or `"agent"` (subagent) | `"command"` |
 | `command` | Shell command to execute | `"$CLAUDE_PROJECT_DIR/.claude/hooks/format.sh"` |
-| `timeout` | Optional timeout in seconds (default 60) | `30` |
+| `timeout` | Optional timeout in seconds. Defaults: 600 for command/http/mcp_tool, 30 for prompt, 60 for agent. | `30` |
 | `once` | If `true`, run the hook only once per session | `true` |
+| `async` | If `true`, runs in the background without blocking | `true` |
+| `asyncRewake` | If `true`, runs in the background and wakes Claude on exit code 2. Implies `async`. | `true` |
+| `shell` | Accepts `"bash"` or `"powershell"`. Defaults to `"bash"`, or to `"powershell"` on Windows when Git Bash isn't installed. | `"bash"` |
+| `statusMessage` | Custom spinner message displayed while the hook runs | `"Formatting…"` |
+
+> **Note**: Some events lower the default timeout. `UserPromptSubmit` lowers the `command`, `http`, and `mcp_tool` default to 30 seconds, and `MessageDisplay` lowers it to 10 seconds. `SessionEnd` hooks share a 1.5-second budget; if your settings set a longer per-hook `timeout`, Claude Code raises that budget to match, up to 60 seconds.

 ### Matcher Patterns

@@ -219,6 +225,8 @@ The hook input (tool name, tool input, session context) is passed as the MCP too

 Subagent-based verification hooks that spawn a dedicated agent to evaluate conditions or perform complex checks. Unlike prompt hooks (single-turn LLM evaluation), agent hooks can use tools and perform multi-step reasoning.

+> **Note**: Agent hooks are experimental and may change.
+
 ```json
 {
  "type": "agent",
@@ -669,6 +677,8 @@ All hooks receive JSON input via stdin:
 > }
 > ```

+> **`retry` (PermissionDenied)**: Use JSON `hookSpecificOutput.retry: true` to tell the model it may retry the denied tool call.
+
 #### `terminalSequence` (v2.1.141)

 Hooks can emit raw OSC (operating system command) escape sequences by setting `terminalSequence` in the JSON output. The host writes the sequence to its controlling terminal when the hook returns — useful for desktop notifications, window-title updates, and terminal bells without requiring a TTY of your own.
@@ -1467,7 +1477,7 @@ echo $?

 | Aspect | Behavior |
 |--------|----------|
-| **Timeout** | 60 seconds default, configurable per command |
+| **Timeout** | 600 seconds default for command/http/mcp_tool (30 for prompt, 60 for agent); configurable per hook |
 | **Parallelization** | All matching hooks run in parallel |
 | **Deduplication** | Identical hook commands deduplicated |
 | **Environment** | Runs in current directory with Claude Code's environment |
@@ -1524,8 +1534,8 @@ Edit `~/.claude/settings.json` or `.claude/settings.json` with the hook configur

 ---

-**Last Updated**: August 15, 2026
-**Claude Code Version**: 2.1.233
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/hooks
 - https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
@@ -87,6 +87,20 @@ Plugin manifest uses JSON format in `.claude-plugin/plugin.json`:
 }
 ```

+Beyond those identity fields, the manifest can point Claude Code at components that live somewhere other than the default folders, and carry discovery and dependency metadata:
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `workflows` | string \| array | Custom [workflow](https://code.claude.com/docs/en/workflows) script files or directories (replaces the default `workflows/`) |
+| `outputStyles` | string \| array | Custom output style files or directories (replaces the default `output-styles/`) |
+| `lspServers` | string \| array \| object | LSP servers for code intelligence — go to definition, find references, diagnostics. Commonly `"./.lsp.json"`. See [LSP server configuration](#lsp-server-configuration) |
+| `channels` | array | Channel declarations for message injection (Telegram, Slack, Discord style) |
+| `dependencies` | array | Other plugins this plugin requires, optionally with semver version constraints |
+| `keywords` | array | Discovery tags used when browsing and searching marketplaces |
+| `metadata` | object | Free-form object for your own data, such as entitlement or catalog fields |
+| `experimental.themes` | string \| array | Color theme files or directories (replaces the default `themes/`) |
+| `experimental.monitors` | string \| array | [Background Monitor](#background-monitors-v21105) configurations that start automatically when the plugin is active |
+
 ## Plugin Structure Example

 ```
@@ -123,13 +137,15 @@ my-plugin/
    └── plugin.test.js
 ```

+> **Note**: `commands/` is **legacy**. Official guidance is *"Use `skills/` for new plugins."* Existing `commands/` directories keep working — the three example plugins in this module ship one — but a new plugin should put its capabilities in `skills/` as `SKILL.md` directories instead of flat Markdown command files.
+
 ### LSP server configuration

 Plugins can include Language Server Protocol (LSP) support for real-time code intelligence. LSP servers provide diagnostics, code navigation, and symbol information as you work.

 **Configuration locations**:
 - `.lsp.json` file in the plugin root directory
- Inline `lsp` key in `plugin.json`
+- The `lspServers` key in `plugin.json` — the official manifest field name. It accepts a string, an array, or an object: a string or array points at LSP config file(s) or directories (for example `"./.lsp.json"`), and an object declares the servers inline.

 #### Field reference

@@ -543,6 +559,12 @@ Example `blockedMarketplaces` with host/path regex (v2.1.119):
 }
 ```

+#### Marketplace `headersHelper` (v2.1.238)
+
+A `url` marketplace — or an individual catalog entry — can name a `headersHelper` command that mints the HTTP headers used to fetch the catalog and any same-origin archives. This is how a private marketplace behind a token-issuing service authenticates without a static secret in the config.
+
+A **catalog entry's** helper runs only on install or update, and only after its command has been shown to you: `claude plugin install` and `claude plugin update` prompt `[y/N]` before running it. Pass `-y` to accept without the prompt in automation.
+
 ### Additional Marketplace Features

 - **Marketplace search bar (v2.1.172)**: When browsing a marketplace's plugins in `/plugin`, a search bar lets you filter the marketplace's plugins by name or keyword — handy for large marketplaces where scrolling the full list is slow.
@@ -620,6 +642,10 @@ Plugins can be sourced from multiple locations:

 GitHub and git sources support optional `ref` (branch/tag) and `sha` (commit hash) fields for version pinning.

+**Bare source names and `metadata.pluginRoot` (v2.1.239)**: a marketplace's `metadata.pluginRoot` now takes effect — a bare plugin source name in the catalog resolves to a directory under that root, instead of having to be spelled as a full relative path on every entry.
+
+**Skills synced from claude.ai (v2.1.239)**: plugins synced down from claude.ai appear as `name@synced`. Address them that way in `claude plugin enable <name>@synced` and `claude plugin disable <name>@synced`. A synced plugin never overrides an installed plugin of the same name — the two coexist, distinguished by the `@synced` suffix.
+
 #### `archive` source (v2.1.224+)

 Install a plugin from a zip over HTTPS — no git clone, no npm install.
@@ -778,8 +804,24 @@ claude plugin tag <version>                  # Create a release git tag with ver
 claude plugin prune                          # Remove orphaned auto-installed plugin dependencies (v2.1.121+)
 claude plugin uninstall <name> --prune       # Uninstall and cascade-clean orphaned dependencies (v2.1.121+)
 claude plugin details <name>                 # Show inventory + projected per-turn token cost (v2.1.139+)
+claude plugin init                           # Scaffold a new plugin (alias: claude plugin new)
 ```

+**Aliases**: `claude plugin new` for `init`, `remove` / `rm` for `uninstall`, `ls` for `list`, and `autoremove` for `prune`.
+
+**Flags worth knowing:**
+
+| Command | Flag | Purpose |
+|---------|------|---------|
+| `plugin init` | `--with <components...>` | Scaffold specific component folders: `skills`, `agents`, `hooks`, `mcp`, `lsp`, `output-style`, `channel` |
+| `plugin init` | `-f`, `--force` | Overwrite an existing `.claude-plugin/` directory |
+| `plugin install` | `--config <key=value>` | Set a `userConfig` option at install time |
+| `plugin install` | `-y`, `--yes` | Accept commands without a confirmation prompt |
+| `plugin list` | `--available` | Also list plugins available from marketplaces (requires `--json`) |
+| `plugin tag` | `--push` | Push the tag to the remote after creating it |
+| `plugin tag` | `--dry-run` | Print what would be tagged without creating the tag |
+| `plugin validate` | `--strict` | Treat warnings as errors |
+
 Example: `claude plugin tag v0.3.0` validates the version format, creates the matching git tag, and is the recommended way to cut plugin releases for distribution.

 `claude plugin prune` is useful after installing or uninstalling marketplace plugins that pulled in their own dependencies — it removes any auto-installed plugins whose parent plugin has since been removed. `plugin uninstall --prune` does the same cascade in a single step.
@@ -1226,8 +1268,8 @@ The following Claude Code features work together with plugins:

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/plugins
 - https://code.claude.com/docs/en/changelog#2-1-172
@@ -294,7 +294,7 @@ Use both together:
 1. **Work normally** - Claude Code creates checkpoints automatically
 2. **Want to go back?** - Press `Esc` twice or use `/rewind`
 3. **Choose checkpoint** - Select from the list to rewind
-4. **Select what to restore** - Choose from restore code and conversation, restore conversation, restore code, summarize from here, or cancel
+4. **Select what to restore** - Choose from restore code and conversation, restore conversation, restore code, summarize from here, summarize up to here, or cancel
 5. **Continue working** - You're back at that point

 ### Keyboard Shortcuts
@@ -336,8 +336,8 @@ Remember: checkpoints are not a replacement for git. Use checkpoints for rapid e

 ---

-**Last Updated**: August 4, 2026
-**Claude Code Version**: 2.1.220
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/checkpointing
 - https://code.claude.com/docs/en/settings
@@ -685,18 +685,15 @@ Claude: [Shows linter output from bg-5002]

 ### Configuration

-```json
-{
-  "backgroundTasks": {
-    "enabled": true,
-    "maxConcurrentTasks": 5,
-    "notifyOnCompletion": true,
-    "autoCleanup": true,
-    "logOutput": true
-  }
-}
+There is no `settings.json` block for background tasks. The feature is controlled by an environment variable:
+
+```bash
+# Turn background tasks off entirely
+export CLAUDE_CODE_DISABLE_BACKGROUND_TASKS=true
 ```

+Concurrency is not a background-task setting either — how many agents run at once is governed by `CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS` (default `20`).
+
 ---

 ## Monitor Tool (Event-Driven Streams)
@@ -1191,7 +1188,7 @@ CLAUDE_CODE_ENABLE_AWAY_SUMMARY=1 claude   # force enable recap

 ## Cross-Session Messaging

-> **Added in v2.1.224**, extended through v2.1.232. Available on **macOS and Linux**.
+> **Added in v2.1.224**, extended through v2.1.239. Available on **macOS, Linux, and (since v2.1.239) Windows**.

 Sessions used to be islands. Cross-session messaging lets one Claude Code session talk to
 another — including sessions on your other machines and your cloud sessions — so you can
@@ -1218,6 +1215,27 @@ SendMessage({ to: "<session name>", message: "What did you conclude about the re
 Since v2.1.232, a bare name is enough — you no longer need to append a disambiguating ref
 unless two rows genuinely share the same name.

+### Waiting for a Session to Go Idle (`notify_when_idle`, v2.1.236)
+
+When the session you are messaging is mid-task, you usually want to know when it finishes
+rather than poll it. `SendMessage` takes a `notify_when_idle` input for exactly that:
+
+```text
+SendMessage({
+  to: "auth-refactor",
+  message: "ping me when the migration finishes",
+  notify_when_idle: true
+})
+```
+
+It is **opt-in and one-shot** — the target session sends a single notice the next time it
+goes idle, and then the subscription is done. There is no polling loop and no repeated
+notification if the session goes busy and idle again.
+
+Two related v2.1.239 changes: `ListAgents` now also reports **the session's own name** (so
+a session can tell others how to address it) alongside its live teammates, and cross-session
+messaging became available on **Windows**.
+
 ### `@`-Mention Shorthand (v2.1.232)

 Instead of calling the tool explicitly, you can `@`-mention a session directly in your
@@ -1248,7 +1266,7 @@ inbound delivery, never loosen it. Since v2.1.232 the setting also has a `/confi
 - Remote Control sessions on your other machines, addressable by name (v2.1.225).
 - A cloud session **receives** your message but cannot message a local session back yet —
  read its answer in its own transcript.
- macOS and Linux only.
+- macOS and Linux from v2.1.224; Windows since v2.1.239.

 ---

@@ -1263,7 +1281,7 @@ Claude Code supports keyboard shortcuts for efficiency. Here's the complete refe
 | `Ctrl+C` | Cancel current input/generation |
 | `Ctrl+D` | Exit Claude Code |
 | `Ctrl+G` | Edit plan in external editor |
-| `Ctrl+L` | Clear terminal screen |
+| `Ctrl+L` | Redraw the screen (repaint only — the double-press `/clear` shortcut was removed in v2.1.238) |
 | `Ctrl+O` | Toggle verbose output (view reasoning) |
 | `Ctrl+R` | Reverse search history. Defaults to **all prompts across all projects** (v2.1.129+); press `Ctrl+S` inside the picker to narrow to the current project. Earlier versions defaulted to project-only. |
 | `Ctrl+T` | Toggle task list view |
@@ -1272,6 +1290,11 @@ Claude Code supports keyboard shortcuts for efficiency. Here's the complete refe
 | `Shift+Tab` / `Alt+M` | Toggle permission modes |
 | `Option+P` / `Alt+P` | Switch model |
 | `Option+T` / `Alt+T` | Toggle extended thinking |
+| `Option+O` / `Alt+O` | Toggle fast mode (`/fast`) |
+| `Ctrl+X` `Ctrl+K` | Stop all background subagents |
+| `Ctrl+S` | Stash the current prompt; press again to restore it |
+| `Ctrl+_` | Undo the last edit to the prompt input |
+| `:` | Type `:` at the start of a word to open emoji shortcode completion, e.g. `:heart:` (v2.1.217+) |

 **Line Editing (standard readline shortcuts):**

@@ -1507,6 +1530,7 @@ For instructions about your project or codebase, use [CLAUDE.md](../02-memory/)
 | **Proactive** | Claude executes immediately and makes reasonable assumptions instead of pausing for routine decisions. Stronger autonomous-execution guidance than auto mode, but it does **not** change your permission mode — you still see permission prompts |
 | **Explanatory** | Adds educational "Insights" between steps, explaining implementation choices and codebase patterns |
 | **Learning** | Collaborative learn-by-doing. Claude shares insights *and* leaves `TODO(human)` markers for you to implement small, strategic pieces yourself |
+| **Concise** (v2.1.237) | Claude leads with the result and skips preamble and narration. Thoroughness is unchanged — only the framing around the answer is dropped. Select it in `/config` → Output style, or set `"outputStyle": "Concise"` |

 ### Selecting a style

@@ -1777,6 +1801,8 @@ Claude Code controls the browser in a visible window — you can watch actions h

 Remote Control lets you continue a locally running Claude Code session from your phone, tablet, or any browser. Your local session keeps running on your machine — nothing moves to the cloud. Available on Pro, Max, Team, and Enterprise plans (v2.1.51+).

+Remote Control is **no longer a research preview** — the label was dropped in week 34 of 2026. Any machine running `claude remote-control` now shows up as a **device card** in the Code tab of the Claude app, so you can start a session on that machine straight from your phone rather than having to start one on the machine first and then connect to it.
+
 ### Starting Remote Control

 **From the CLI**:
@@ -1982,7 +2008,7 @@ Connect external services for richer context:

 ### Permission modes in Desktop

-The Desktop App supports the same 4 permission modes as the CLI:
+The Desktop App supports the same permission modes as the CLI:

 | Mode | Behavior |
 |------|----------|
@@ -2187,6 +2213,8 @@ Example of `deniedDomains` overriding a broad wildcard (v2.1.113+):

 The wildcard lets everything on `example.com` through, but `deniedDomains` still blocks the specifically-named host.

+> **Note** (v2.1.243): the sandboxed Bash tool's permission prompt **no longer lists the allowed network hosts**. Claude simply attempts the request, and you approve each new host as it comes up — so do not expect the prompt to show you the allowlist up front. The same release also stopped dropping network-violation details when the blocked command happens to exit `0`, so a silent-looking success now still reports what was blocked.
+
 ### Credential Masking (v2.1.221, v2.1.224)

 > **Changelog-sourced**: these `sandbox.credentials` options come from the v2.1.221 and
@@ -2389,8 +2417,17 @@ These keys go in `~/.claude/settings.json` (or a project `.claude/settings.json`
 | `enableArtifact` | Per-user enable/disable of the Artifact tool (v2.1.196). |
 | `crossSessionInbound` | (v2.1.224) How inbound [cross-session messages](#cross-session-messaging) are handled — `"accept"`, `"hold"`, or `"refuse"`. Project and local values apply only when *stricter* on the `accept < hold < refuse` ladder. Exposed in `/config` as "Messages from your other sessions" since v2.1.232. |
 | `dialogExpiry` | (v2.1.224) How long an unanswered dialog stays open. Default `"5m"`; accepts `"60s"`, `"5m"`, `"10m"`, or `"never"`. Overridden by `CLAUDE_CODE_USER_DIALOG_TIMEOUT_MS`. Exposed in `/config` as "Dialog expiry" since v2.1.232. |
+| `modelPicker` | (v2.1.243) Choose which models the `/model` picker lists, in your own order and with your own labels. One of the few settings that **replaces rather than merges** across settings layers — the nearest-scope value wins outright. |
+| `promptCacheTtl` | (v2.1.243) Choose the prompt cache lifetime for the main conversation. |
+| `subagentPromptCacheTtl` | (v2.1.243) The same choice for subagents and other requests outside the main conversation. |
+| `modelPricing` | (v2.1.243) **Managed setting.** Supplies your organization's contracted rates so `/cost`, the status line, and telemetry report those instead of list price. |
+| `keybindingFlavor` | (v2.1.238) `"classic"` (default) or `"readline"`. `readline` makes `Ctrl+W` delete back to the previous whitespace, as Bash does; v2.1.239 extended it to `Alt+F`, `Ctrl`/`Option+→`, and `Alt+D`. |
 | `spellcheck` | (v2.1.235) Underlines misspelled words in the prompt input using whichever of `aspell`, `hunspell`, or `ispell` is on your `PATH`, tried in that order. Object-valued — `{"enabled": true, "language": "en_GB"}` — and off by default. **Read from user settings, the `--settings` flag, and managed settings only**: a `spellcheck` block in a project `.claude/settings.json` or `.claude/settings.local.json` is ignored. |

+> **Changelog-sourced**: `modelPricing` is documented from the v2.1.243 changelog entry;
+> the settings reference does not yet list it. The other four keys above appear in the
+> official settings reference.
+
 ### Fallback Models (`fallbackModel`)

 The `fallbackModel` setting lets you configure **up to three** fallback models, tried in order, when the primary model is overloaded or unavailable.
@@ -2410,6 +2447,7 @@ Override config with environment variables:
 ```bash
 # Model selection
 export ANTHROPIC_MODEL=claude-opus-4-8
+export ANTHROPIC_DEFAULT_MODEL=claude-opus-4-8   # (v2.1.236) Model new sessions start on. Unlike ANTHROPIC_MODEL, a /model pick still overrides it — and that pick persists across restarts
 export ANTHROPIC_DEFAULT_OPUS_MODEL=claude-opus-4-8
 export ANTHROPIC_DEFAULT_SONNET_MODEL=claude-sonnet-4-6
 export ANTHROPIC_DEFAULT_HAIKU_MODEL=claude-haiku-4-5
@@ -2691,8 +2729,8 @@ For more information about Claude Code and related features:

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/settings
 - https://code.claude.com/docs/en/sandboxing
@@ -2702,4 +2740,6 @@ For more information about Claude Code and related features:
 - https://code.claude.com/docs/en/model-config
 - https://code.claude.com/docs/en/permission-modes
 - https://code.claude.com/docs/en/settings.md
+- https://code.claude.com/docs/en/settings-reference
+- https://code.claude.com/docs/en/whats-new/2026-w34
 **Compatible Models**: Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -399,6 +399,8 @@ claude --ide "help me with this file"

 The `--agents` flag accepts a JSON object defining custom subagents for a session.

+As of **v2.1.243**, `--agents` no longer silently ignores invalid JSON or an invalid agent definition — it exits with a clear error, matching how `--mcp-config` already behaved.
+
 ### Agents JSON Format

 ```json
@@ -811,6 +813,7 @@ The "ultrathink" keyword in prompts activates deep reasoning. The `/effort` menu
 |----------|-------------|
 | `ANTHROPIC_API_KEY` | API key for authentication |
 | `ANTHROPIC_MODEL` | Override default model |
+| `ANTHROPIC_DEFAULT_MODEL` | (v2.1.236) Sets the model new sessions start on. Unlike `ANTHROPIC_MODEL`, which pins the model, a `/model` pick still overrides this value **and persists across restarts** — that contrast is the point of the variable. |
 | `ANTHROPIC_CUSTOM_MODEL_OPTION` | Custom model option for API |
 | `ANTHROPIC_DEFAULT_OPUS_MODEL` | Override default Opus model ID |
 | `ANTHROPIC_DEFAULT_SONNET_MODEL` | Override default Sonnet model ID |
@@ -901,6 +904,11 @@ These keys live in a `settings.json` file (`~/.claude/settings.json` for user sc
 | `emojiCompletionEnabled` | (v2.1.217) Enables emoji shortcode autocomplete in the prompt input (e.g. typing `:heart:` inserts ❤️). Set `false` to disable. |
 | `workflowSizeGuideline` | (v2.1.219) Sets the advisory Dynamic workflow size guideline from any settings file. The guideline is guidance Claude aims for, not a hard cap — the default is medium (aim for fewer than 15 agents), and other sizes or unrestricted can be selected. While this key is set, the "Dynamic workflow size" row is hidden in `/config`. Distinct from `CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS`, which is an enforced concurrency limit. |
 | `spellcheck` | (v2.1.235) Underlines misspelled words in the prompt input using whichever of `aspell`, `hunspell`, or `ispell` is on your `PATH`, tried in that order. Object-valued — `{"enabled": true, "language": "en_GB"}` — and off by default. **Read from user settings, the `--settings` flag, and managed settings only**: a `spellcheck` block in a project `.claude/settings.json` or `.claude/settings.local.json` is ignored. See also [Advanced Features → Additional Per-User Settings](../09-advanced-features/README.md#additional-per-user-settings). |
+| `modelPicker` | (v2.1.243) Choose which models the `/model` picker lists, in your own order and with your own labels. One of the few settings that **replaces rather than merges** across settings layers. |
+| `promptCacheTtl` | (v2.1.243) Choose the prompt cache lifetime for the main conversation. |
+| `subagentPromptCacheTtl` | (v2.1.243) The same choice for subagents and other requests outside the main conversation. |
+| `modelPricing` | (v2.1.243) **Managed setting.** Supplies your organization's contracted rates so `/cost`, the status line, and telemetry report those instead of list price. **Changelog-sourced** — the settings reference does not yet list this key. |
+| `keybindingFlavor` | (v2.1.238) `"classic"` (default) or `"readline"`. `readline` makes `Ctrl+W` delete back to the previous whitespace, as Bash does; v2.1.239 extended it to `Alt+F`, `Ctrl`/`Option+→`, and `Alt+D`. |

 ```json
 {
@@ -1014,8 +1022,8 @@ claude -p --output-format json "query"

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/cli-reference
 - https://code.claude.com/docs/en/env-vars
@@ -1039,4 +1047,5 @@ claude -p --output-format json "query"
 - https://code.claude.com/docs/en/headless
 - https://code.claude.com/docs/en/cli-reference.md
 - https://code.claude.com/docs/en/settings.md
+- https://code.claude.com/docs/en/settings-reference
 **Compatible Models**: Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -91,7 +91,7 @@ Commands are user-invoked shortcuts that execute specific actions.
 | `/stickers` | View session stickers | Fun rewards |
 | `/fast` | Toggle fast output mode; applies to **Opus 5 and Opus 4.8** (v2.1.219) | Speed up responses |
 | `/terminal-setup` | Configure terminal integration | Setup terminal features |
-| `/undo` | Alias for `/rewind` (v2.1.108) | Same as `/rewind` |
+| `/undo` | **No longer documented** — added as an alias for `/rewind` in v2.1.108, but it appears nowhere in the official commands reference | Use `/rewind` (or `Esc Esc`) instead |
 | `/upgrade` | Check for updates | Version management |
 | `/team-onboarding` | Generate a teammate ramp-up guide from this project's Claude Code usage | Onboarding new teammates (v2.1.101) |
 | `/code-review ultra` | Run a cloud multi-agent code review over your current changes. `/ultrareview` remains as an alias; `/code-review ultra` is the preferred invocation. Includes 3 free runs on Pro and Max, then requires usage credits | Deep pre-merge review across multiple agents (v2.1.112) |
@@ -152,10 +152,10 @@ Specialized AI assistants with isolated contexts for specific tasks.
 |-------|-------------|-------|-------|-------------|
 | **general-purpose** | Multi-step tasks, research | All tools | Inherits model | Complex research, multi-file tasks |
 | **Plan** | Implementation planning | Read, Glob, Grep, Bash | Inherits model | Architecture design, planning |
-| **Explore** | Codebase exploration | Read, Glob, Grep | Haiku 4.5 | Quick searches, understanding code |
-| **Bash** | Command execution | Bash | Inherits model | Git operations, terminal tasks |
+| **Explore** | Codebase exploration | Read, Glob, Grep | Inherits (capped at Opus) | Quick searches, understanding code |
+| **claude** | Catch-all for tasks that don't fit a more specialized agent | All tools | Inherits model | Tasks with no specialized agent; default agent for a dispatched background session |
 | **statusline-setup** | Status line configuration | Bash, Read, Write | Sonnet 4.6 | Configure status line display |
-| **Claude Code Guide** | Help and documentation | Read, Glob, Grep | Haiku 4.5 | Getting help, learning features |
+| **claude-code-guide** | Help and documentation | Read, Glob, Grep | Haiku 4.5 | Getting help, learning features |

 ### Subagent Configuration Fields

@@ -246,6 +246,7 @@ cp -r 03-skills/* ~/.claude/skills/
 | `/batch` | Run prompts on multiple files | Batch operations |
 | `/claude-api` | Build apps with Claude API | API development |
 | `/debug` | Debug failing tests/errors | Debugging sessions |
+| `/design` *(research preview, v2.1.233+)* | Create a multi-artboard design canvas — UI mockups, screen flows, landing pages, posters — refined visually instead of in code. Pro/Max/Team/Enterprise | Designing a screen or page you would rather tweak by hand than in HTML |
 | `/fewer-permission-prompts` | Scan transcripts and propose a prioritized allowlist | Reduce repeat permission prompts |
 | `/loop` | Run prompts on interval | Recurring tasks |
 | `/run` *(v2.1.145+)* | Launch this project's app to see a change running | Verifying a change in the real app |
@@ -462,13 +463,13 @@ cp 02-memory/personal-CLAUDE.md ~/.claude/CLAUDE.md
 | **/proactive** | Alias for `/loop` — same recurring-task behavior (v2.1.105) | Use `/proactive` interchangeably with `/loop` |
 | **/recap** | Show a session recap when returning to an existing session (v2.1.108) | Run `/recap` after being away to get context on what was done |
 | **/tui** | Toggle fullscreen TUI (text user interface) mode for flicker-free rendering (v2.1.110) | Use `/tui` in fullscreen terminals or tmux |
-| **/undo** | Alias for `/rewind` — reverts to the previous checkpoint (v2.1.108) | Use `/undo` interchangeably with `/rewind` |
+| **/undo** | **No longer documented** — added as an alias for `/rewind` in v2.1.108, but it appears nowhere in the official commands reference | Use `/rewind` (or `Esc Esc`) instead |
 | **Monitor Tool** | Watch a background command's stdout stream and react to events instead of polling (v2.1.98+) | Use the Monitor tool via [Advanced Features](09-advanced-features/) |
-| **Output Styles** | Change Claude's role, tone, and default response format via the system prompt. Built-ins: Default, Proactive, Explanatory, Learning | `/config` → Output style, or set `outputStyle`. The `/output-style` command was removed in v2.1.91. See [Advanced Features](09-advanced-features/#output-styles) |
+| **Output Styles** | Change Claude's role, tone, and default response format via the system prompt. Built-ins: Default, Proactive, Explanatory, Learning, Concise | `/config` → Output style, or set `outputStyle`. The `/output-style` command was removed in v2.1.91. See [Advanced Features](09-advanced-features/#output-styles) |
 | **Status Line** | Render a custom status line from a command that receives session, model, cost, and context JSON on stdin | `/statusline` or the `statusLine` setting. See [Advanced Features](09-advanced-features/#status-line) |
 | **Community Marketplace** | Third-party plugins that passed Anthropic's automated validation, each pinned to a commit SHA | `/plugin marketplace add anthropics/claude-plugins-community`, then `/plugin install <name>@claude-community` |
 | **/team-onboarding** | Auto-generate a teammate ramp-up guide from the project's Claude Code setup (v2.1.101) | Run `/team-onboarding` in your project |
-| **Remote Control** | Control Claude Code sessions remotely via API | Use the remote control API to send prompts and receive responses programmatically |
+| **Remote Control** | Control Claude Code sessions remotely via API. **No longer a research preview** — any machine running `claude remote-control` shows up as a device card in the Claude app's Code tab, so a session can be started on it from a phone | Run `claude remote-control` on the machine, then pick its device card in the Claude app. See [Advanced Features](09-advanced-features/README.md) |
 | **Web Sessions** | Run Claude Code in a browser-based environment | Access via `claude web` or through the Anthropic Console |
 | **Desktop App** | Native desktop application for Claude Code | Use `/desktop` or download from Anthropic website |
 | **Cross-Session Messaging** | Sessions can message each other — including your other machines and cloud sessions — discovered via `ListAgents` (v2.1.224+, macOS/Linux) | See [Advanced Features](09-advanced-features/README.md#cross-session-messaging); control inbound with `crossSessionInbound` |
@@ -503,6 +504,9 @@ cp 02-memory/personal-CLAUDE.md ~/.claude/CLAUDE.md
 | **Usage-Limit Auto-Continue** | Claude Code auto-continues a session when a claude.ai usage limit resets, if it was blocked on that limit (v2.1.234) | On by default; turn it off in `/config` → "Continue automatically at usage limit". See [Advanced Features](09-advanced-features/) |
 | **`spellcheck` setting** | Underlines misspelled words in the prompt input using whichever of aspell, hunspell, or ispell is on your `PATH`; off by default (v2.1.235) | Install a checker, then set `"spellcheck": { "enabled": true }` in `~/.claude/settings.json`. User, `--settings`, and managed settings only — ignored in project settings. See [Advanced Features](09-advanced-features/) |
 | **Agent Teams Default Model** | The "Default teammate model" `/config` setting was removed; teammates now inherit the team lead's model by default unless the spawn call specifies one explicitly (v2.1.234) | See [Subagents — Agent Teams](04-subagents/README.md#agent-teams-experimental) |
+| **`/design`** | Design canvas — a multi-artboard visual design (UI mockups, screen flows, landing pages, posters) built on artifacts and refined visually rather than in code. Research preview; requires v2.1.233+; Pro/Max/Team/Enterprise | Run `/design` in the CLI or the Desktop app. See [Slash Commands](01-slash-commands/README.md) |
+| **`notify_when_idle`** | Cross-session `SendMessage` input that asks another session on the same machine to send one notice when it next goes idle — opt-in, one-shot, no polling (v2.1.236). Related: `ListAgents` reports the session's own name and lists live teammates, and Windows gained cross-session messaging (v2.1.239) | Pass `notify_when_idle` to `SendMessage`. See [Advanced Features](09-advanced-features/README.md#cross-session-messaging) |
+| **Plugin manifest fields** | `plugin.json` accepts `workflows`, `channels`, `dependencies` (semver), `outputStyles`, `keywords`, `metadata`, `lspServers`, and `experimental.themes` / `experimental.monitors`. CLI gained `claude plugin new`, `remove`/`rm`, `prune`/`autoremove`, and the flags `--with`, `-f`/`--force`, `--available`, `--push`, `--dry-run` | See [Plugins](07-plugins/README.md) |

 ---

@@ -562,8 +566,8 @@ chmod +x ~/.claude/hooks/*.sh

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/sub-agents
 - https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
@@ -1,5 +1,56 @@
 # Changelog

+## [v2.1.245] — 2026-08-25
+
+### Documentation sync against Claude Code v2.1.245
+
+Claude Code moved from v2.1.235 (2026-08-19) to **v2.1.245** (2026-08-25) — ten
+releases. This entry records that sync.
+
+> **Note**: This changelog has no entries for the v2.1.233 and v2.1.235 syncs
+> (commits `deb7e5c` and `da6e09e`), which landed in the repository without a
+> changelog entry. That gap is recorded here rather than back-filled with
+> reconstructed dates.
+
+### Fixed
+
+- **Version-currency claims were stale and mutually contradictory** — the root
+  `README.md` said `v2.1.235` in one place and `v2.1.220 (July 2026)` in
+  another, and this changelog's top entry still called v2.1.220 the current
+  Claude Code release. All now say v2.1.245.
+- **Hook default timeout** corrected from 60 seconds to **600** for
+  `command`, `http`, and `mcp_tool` (30 for `prompt`, 60 for `agent`), with the
+  per-event overrides documented.
+- **Skill precedence** corrected to **enterprise > personal > project**; the
+  lesson had project and personal swapped, contradicting its own quiz.
+- **Memory files are concatenated, not overridden** — the last two
+  "overrides root CLAUDE.md" strings in `02-memory/README.md` were replaced.
+- **`SECURITY.md`** no longer advertises support for "1.x releases".
+- **`STYLE_GUIDE.md`** no longer teaches a stale footer by example.
+- Graded material (the `lesson-quiz` and `self-assessment` skills) was realigned
+  with the lessons it grades: 31 hook events, five hook types, six rewind
+  options, six permission modes, import depth 4.
+
+### Added
+
+- **WebSocket (`ws`) MCP transport** — a fourth transport, configured through
+  `.mcp.json` or `claude mcp add-json`.
+- **`Concise` output style** — a fifth built-in (v2.1.237).
+- **New settings keys** — `modelPicker`, `promptCacheTtl`,
+  `subagentPromptCacheTtl`, `modelPricing`, `keybindingFlavor`.
+- **`ANTHROPIC_DEFAULT_MODEL`** env var (v2.1.236).
+- **`/design`** — the design-canvas research preview.
+- **`notify_when_idle`** on cross-session `SendMessage` (v2.1.236).
+- **Plugin manifest fields** — `workflows`, `channels`, `dependencies`, and
+  related CLI additions.
+
+### Changed
+
+- **Remote Control** left research preview; machines running
+  `claude remote-control` now appear as device cards in the Claude app.
+- **`Ctrl+L`** repaints the screen rather than clearing it (v2.1.238).
+- **Plugin `commands/`** is marked legacy; `skills/` is the current guidance.
+
 ## [v2.1.220-r2] — 2026-08-04

 ### Accuracy pass against Claude Code v2.1.220 (no upstream version change)
@@ -1166,3 +1217,11 @@ The `vi/`, `zh/`, and `uk/` localized trees are community-maintained and may lag
 - Update all "Last Updated" dates to February 2026

 **Full Changelog**: https://github.com/luongnv89/claude-howto/compare/20779db...v2.0.0
+
+---
+
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
+**Sources**:
+- https://code.claude.com/docs/en/changelog
+**Compatible Models**: Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -377,7 +377,8 @@ Advanced capabilities for complex workflows.
 | Dynamic Workflows | Deterministic multi-agent orchestration via `/workflows` (v2.1.154) | Comprehensive audits, migrations, scale-out |
 | Scheduled Tasks | Recurring tasks with `/loop` and cron tools | Automated recurring workflows |
 | Chrome Integration | Browser automation via headless Chromium | Web testing and scraping |
-| Remote Control (expanded) | Connection methods, security, comparison table | Remote session management |
+| Remote Control (expanded) | Connection methods, security, comparison table, device cards | Remote session management (no longer a research preview) |
+| Cross-Session Messaging | `SendMessage` / `ListAgents`, including `notify_when_idle` (v2.1.236) | Coordinating sessions on the same machine |
 | Keyboard Customization | Custom keybindings, chord support, contexts | Personalized shortcuts |
 | Desktop App (expanded) | Connectors, launch.json, enterprise features | Desktop integration |
 | | | |
@@ -448,6 +449,12 @@ Advanced capabilities for complex workflows.
 - Connection methods and protocols
 - Security considerations and best practices
 - Comparison table of remote access options
+- Out of research preview — a machine running `claude remote-control` appears as a device card in the Claude app's Code tab
+
+### Cross-Session Messaging
+- `SendMessage` between sessions on the same machine
+- `notify_when_idle` — one opt-in notice when the other session next goes idle (v2.1.236)
+- `ListAgents` reports the session's own name and lists live teammates (v2.1.239)

 ### Keyboard Customization
 - Custom keybindings configuration
@@ -888,8 +895,8 @@ Want to add more examples? Follow the structure:

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/overview
 - https://code.claude.com/docs/en/hooks
@@ -372,7 +372,7 @@ Before starting Level 3, make sure you're comfortable with these Level 2 concept
 ✅ Auto Mode with background safety classifier
 ✅ Channels for structured multi-session workflows
 ✅ Voice Dictation for hands-free interaction
-✅ Remote control, desktop app, and web sessions
+✅ Remote control (generally available — your machine appears as a device card in the Claude app), desktop app, and web sessions
 ✅ Agent Teams for multi-agent collaboration

 #### Hands-on Exercises
@@ -424,6 +424,7 @@ export CLAUDE_AGENT_TEAMS=1
 - [ ] Explored Channels for multi-session workflows
 - [ ] Tried Voice Dictation for hands-free input
 - [ ] Understand Remote Control, Desktop App, and Web sessions
+- [ ] Used `notify_when_idle` on a cross-session `SendMessage` to be told when another session goes idle
 - [ ] Enabled and used Agent Teams for collaborative tasks
 - [ ] Used `/loop` for recurring tasks or scheduled monitoring

@@ -697,20 +698,21 @@ Use these checklists to track your progress by level. Run `/self-assessment` any
 Once you've completed all milestones:

 1. **Create team documentation** - Document your team's Claude Code setup
-2. **Build custom plugins** - Package your team's workflows
-3. **Explore Remote Control** - Control Claude Code sessions programmatically from external tools
-4. **Try Web Sessions** - Use Claude Code through browser-based interfaces for remote development
-5. **Use the Desktop App** - Access Claude Code features through the native desktop application
-6. **Use Auto Mode** - Let Claude work autonomously with a background safety classifier
-7. **Leverage Auto Memory** - Let Claude learn your preferences automatically over time
-8. **Set up Agent Teams** - Coordinate multiple agents on complex, multi-faceted tasks
-9. **Use Channels** - Organize work across structured multi-session workflows
-10. **Try Voice Dictation** - Use hands-free voice input for interaction with Claude Code
-11. **Use Scheduled Tasks** - Automate recurring checks with `/loop` and cron tools
-12. **Contribute examples** - Share with the community
-13. **Mentor others** - Help teammates learn
-14. **Optimize workflows** - Continuously improve based on usage
-15. **Stay updated** - Follow Claude Code releases and new features
+2. **Build custom plugins** - Package your team's workflows, including the newer manifest fields (`workflows`, `channels`, `dependencies`)
+3. **Try `/design`** - Sketch a UI mockup, screen flow, or landing page on a design canvas instead of hand-writing HTML (research preview, v2.1.233+)
+4. **Explore Remote Control** - Control Claude Code sessions programmatically from external tools, or start one from your phone via its device card in the Claude app
+5. **Try Web Sessions** - Use Claude Code through browser-based interfaces for remote development
+6. **Use the Desktop App** - Access Claude Code features through the native desktop application
+7. **Use Auto Mode** - Let Claude work autonomously with a background safety classifier
+8. **Leverage Auto Memory** - Let Claude learn your preferences automatically over time
+9. **Set up Agent Teams** - Coordinate multiple agents on complex, multi-faceted tasks
+10. **Use Channels** - Organize work across structured multi-session workflows
+11. **Try Voice Dictation** - Use hands-free voice input for interaction with Claude Code
+12. **Use Scheduled Tasks** - Automate recurring checks with `/loop` and cron tools
+13. **Contribute examples** - Share with the community
+14. **Mentor others** - Help teammates learn
+15. **Optimize workflows** - Continuously improve based on usage
+16. **Stay updated** - Follow Claude Code releases and new features

 ---

@@ -738,8 +740,8 @@ Once you've completed all milestones:

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/overview
 - https://code.claude.com/docs/en/hooks
@@ -390,7 +390,7 @@ cp -r 03-skills/code-review-specialist ~/.claude/skills/
 | **Auto Mode** | Fully autonomous operation with background classifier; available by default on Bedrock/Vertex/Foundry as of v2.1.207 | `Shift+Tab` to cycle modes, or `--permission-mode auto` |
 | **Channels** | Discord and Telegram integration | `--channels` flag, Discord/Telegram bots |
 | **Voice Dictation** | Speak commands and context to Claude | `/voice` command |
-| **Output Styles** | Change Claude's role, tone, and default response format | `/config` → Output style, or `outputStyle` setting. Built-ins: Default, Proactive, Explanatory, Learning |
+| **Output Styles** | Change Claude's role, tone, and default response format | `/config` → Output style, or `outputStyle` setting. Built-ins: Default, Proactive, Explanatory, Learning, Concise |
 | **Status Line** | Custom bottom-of-session status from a command | `/statusline`, or `statusLine` setting; receives session/model/cost/context JSON on stdin |
 | **Hooks (31 events)** | Expanded hook system with 5 types | command, http, mcp_tool, prompt, agent hook types |
 | **MCP Elicitation** | MCP servers can request user input at runtime | Auto-prompted when server needs clarification |
@@ -516,9 +516,10 @@ Getting started checklist:

 ---

-**Last Updated**: August 15, 2026
-**Claude Code Version**: 2.1.233
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
+- https://code.claude.com/docs/en/changelog
 - https://code.claude.com/docs/en/cli-reference
 - https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
 - https://code.claude.com/docs/en/sub-agents
@@ -102,7 +102,7 @@ Run `/lesson-quiz [topic]` after each module. The quiz pinpoints what you missed

 - **GitHub stars** from developers who use Claude Code daily
 - **Forks** from teams adapting this guide for their own workflows
- **Actively maintained** — synced with every Claude Code release (latest: v2.1.235, August 2026)
+- **Actively maintained** — synced with every Claude Code release (latest: v2.1.245, August 2026)
 - **Community-driven** — contributions from developers who share their real-world configurations

 [![Star History Chart](https://api.star-history.com/svg?repos=luongnv89/claude-howto&type=Date)](https://star-history.com/#luongnv89/claude-howto&Date)
@@ -201,7 +201,7 @@ cp -r 03-skills/code-review-specialist ~/.claude/skills/
 Yes. MIT licensed, free forever. Use it in personal projects, at work, in your team — no restrictions beyond including the license notice.

 **Is this maintained?**
-Actively. The guide is synced with every Claude Code release. Current version: v2.1.220 (July 2026), compatible with Claude Code 2.1+.
+Actively. The guide is synced with every Claude Code release. Current version: v2.1.245 (August 2026), compatible with Claude Code 2.1+.

 **How is this different from the official docs?**
 The official docs are a feature reference. This guide is a tutorial with diagrams, production-ready templates, and a progressive learning path. They complement each other — start here to learn, reference the docs when you need specifics.
@@ -873,8 +873,8 @@ MIT License - see [LICENSE](LICENSE). Free to use, modify, and distribute. The o

 ---

-**Last Updated**: August 19, 2026
-**Claude Code Version**: 2.1.235
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/hooks
 - https://code.claude.com/docs/en/overview
@@ -11,7 +11,6 @@ We provide security updates for the following versions:
 | Version | Status | Support Until |
 |---------|--------|---------------|
 | Latest (main) | ✅ Active | Current + 6 months |
-| 1.x releases | ✅ Active | Until next major version |

 **Note**: As an educational guide project, we focus on maintaining current best practices and documentation security rather than traditional version support. Updates are applied directly to the main branch.

@@ -326,8 +325,8 @@ This security policy is reviewed and updated:
 - When the project scope changes
 - Annually as a minimum

-**Last Updated**: August 4, 2026
-**Claude Code Version**: 2.1.220
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/overview
 **Compatible Models**: Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -597,11 +597,12 @@ Lesson READMEs end with a metadata block:

 ```markdown
 ---
-**Last Updated**: July 29, 2026
-**Claude Code Version**: 2.1.220
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Compatible Models**: Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
 ```

+- Use the version captured by the current sync run, not the value shown here
 - Use month + day + year format (e.g., "May 20, 2026")
 - Update the version when features change
 - List all compatible models
@@ -631,8 +632,8 @@ Before submitting content, verify:

 ---

-**Last Updated**: July 29, 2026
-**Claude Code Version**: 2.1.220
+**Last Updated**: August 25, 2026
+**Claude Code Version**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/overview
 - https://code.claude.com/docs/en/changelog
@@ -1433,16 +1433,14 @@ Complete configuration example:
    "enabled": true,
    "showThinkingProcess": true
  },
-  "backgroundTasks": {
-    "enabled": true,
-    "maxConcurrentTasks": 5
-  },
  "permissions": {
    "defaultMode": "manual"
  }
 }
 ```

+There is no `settings.json` block for background tasks — the feature is controlled by the `CLAUDE_CODE_DISABLE_BACKGROUND_TASKS` environment variable, and concurrency by `CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS` (default `20`).
+
 **See**: [09-advanced-features/](09-advanced-features/) for comprehensive guide

 ---
@@ -93,7 +93,6 @@
 | `/theme` | テーマピッカーを開く／カスタムテーマを管理（v2.1.118）。カスタムテーマは `~/.claude/themes/<name>.json` に JSON で定義 |
 | `/tui` | ちらつきのないフルスクリーン TUI（テキストユーザーインターフェース）モードを切り替え（v2.1.110 で追加） |
 | `/ultrareview` | マルチエージェント分析による包括的なクラウドベースのコードレビュー（v2.1.111 で追加）。現在の推奨呼び出しは `/code-review ultra` で、`/ultrareview` はそのエイリアスとして残っている。Pro と Max では 3 回まで無料、以降は使用クレジットが必要 |
-| `/undo` | `/rewind` のエイリアス（v2.1.108 で追加） |
 | `/upgrade` | 上位プランへのアップグレードページを開く |
 | `/usage` | 正式な使用状況ダッシュボード（v2.1.118）— プラン使用量制限、レート制限、コスト、日次セッション統計を統合。`/cost` と `/stats` は特定のタブを開くタイピング用エイリアス |
 | `/voice` | プッシュトゥトーク音声入力を切り替え |
@@ -118,6 +117,7 @@
 | `/fork` | `/branch` に改名（エイリアスは引き続き有効、v2.1.77） |
 | `/pr-comments` | v2.1.91 で削除 — Claude に直接 PR コメントを見るよう依頼する |
 | `/vim` | v2.1.92 で削除 — /config → エディタモードを使う |
+| `/undo` | v2.1.245 時点で公式コマンドリファレンスに掲載されていない（v2.1.108 で `/rewind` のエイリアスとして追加された） — `/rewind` または `Esc` 二度押しを使う |

 ### 直近の変更点

@@ -140,7 +140,7 @@
 - ちらつきのないフルスクリーン TUI 描画用に `/tui` コマンドを追加（v2.1.110）
 - フォーカスビュー切替用に `/focus` コマンドを追加。`Ctrl+O` は詳細トランスクリプトの切替のみに（v2.1.110）
 - セッションコンテキストのおさらいを手動で発火する `/recap` コマンドを追加（v2.1.108）
- `/undo` を `/rewind` のエイリアスとして追加（v2.1.108）
+- `/undo` を `/rewind` のエイリアスとして追加（v2.1.108）。v2.1.245 時点で公式コマンドリファレンスには掲載されていない — `/rewind` または `Esc` 二度押しを使う
 - `/proactive` を `/loop` のエイリアスとして追加（v2.1.105）
 - `/effort` がインタラクティブな矢印キースライダーを獲得し、`high` と `max` の間に新しい `xhigh` レベルを追加。Opus 4.7 プランではデフォルト思考量が `xhigh` に引き上げられた（v2.1.111）
 - 包括的なクラウドベースのマルチエージェントコードレビュー用に `/ultrareview` を追加（v2.1.111）。v2.1.223 で `/code-review ultra` が推奨の呼び出しとなり、`/ultrareview` はエイリアスとして残った
@@ -603,8 +603,8 @@ Deploy the application to production:

 ---

-**Last Updated**: August 15, 2026
-**Claude Code Version**: 2.1.233
+**最終更新**: 2026 年 8 月 25 日
+**Claude Code バージョン**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/commands
 - https://code.claude.com/docs/en/interactive-mode
@@ -685,7 +685,9 @@ Claude は、現在の作業ディレクトリのメモリファイルに加え
 ````markdown
 # API Module Standards

-This file overrides root CLAUDE.md for everything in /src/api/
+This file supplements root CLAUDE.md for everything in /src/api/. Memory files are
+concatenated, not overridden — the root CLAUDE.md still applies, and Claude Code
+loads this file on demand when it reads files in this subtree.

 ## API-Specific Standards

@@ -1127,7 +1129,9 @@ Claude はどのメモリファイルを更新するかを尋ねる。
   cat > /path/to/directory/CLAUDE.md << 'EOF'
   # [Directory Name] Standards

-   This file overrides root CLAUDE.md for this directory.
+   This file supplements root CLAUDE.md for this directory. Memory files are
+   concatenated, not overridden — Claude Code loads this file on demand when it
+   reads files in this directory.

   ## [Specific Standards]
   EOF
@@ -1202,8 +1206,8 @@ Claude はどのメモリファイルを更新するかを尋ねる。
 - [公式メモリドキュメント](https://code.claude.com/docs/en/memory) — Anthropic ドキュメント

 ---
-**Last Updated**: April 24, 2026
-**Claude Code Version**: 2.1.119
+**最終更新**: 2026 年 8 月 25 日
+**Claude Code バージョン**: 2.1.245
 **Sources**:
 - https://code.claude.com/docs/en/memory
 - https://code.claude.com/docs/en/settings
@@ -96,7 +96,7 @@ sequenceDiagram
 | **Project** | `.claude/skills/<skill-name>/SKILL.md` | チーム | あり（git 経由） | チームの標準 |
 | **Plugin** | `<plugin>/skills/<skill-name>/SKILL.md` | 有効化された範囲 | プラグインに依存 | プラグインへのバンドル |

-スキル名がレベル間で重複する場合、優先度の高い配置場所が優先される: **enterprise > project > personal**。プロジェクトスキルはデフォルトで個人スキルを上書きする。プラグインスキルは `plugin-name:skill-name` の名前空間を用いるため衝突しない。
+スキル名がレベル間で重複する場合、優先度の高い配置場所が優先される: **enterprise > personal > project**。個人スキルはデフォルトでプロジェクトスキルを上書きする。プラグインスキルは `plugin-name:skill-name` の名前空間を用いるため衝突しない。

 ### 自動検出

@@ -140,7 +140,7 @@ Show concrete examples of using this Skill.
 ### 必須フィールド

 - **name**: 小文字、数字、ハイフンのみ（最大 64 文字）。"anthropic" や "claude" を含めることはできない。
- **description**: スキルが何をするかと、いつ使うか（最大 1024 文字）。Claude がスキルを発火させるべきタイミングを判断する上で重要である。
+- **description**: スキルが何をするかと、いつ使うか。`description` と `when_to_use` を合わせたテキストはスキル一覧で **1,536 文字**で切り詰められる（`skillListingMaxDescChars` で変更可能）。Claude がスキルを発火させるべきタイミングを判断する上で重要である。

 ### オプションのフロントマターフィールド

@@ -170,7 +170,8 @@ paths: "src/api/**/*.ts"               # スキルの発火を制限する glob
 | フィールド | 説明 |
 |-------|-------------|
 | `name` | 小文字、数字、ハイフンのみ（最大 64 文字）。"anthropic" や "claude" を含められない。 |
-| `description` | スキルが何をするかと、いつ使うか（最大 1024 文字）。自動呼び出しのマッチングに重要。 |
+| `description` | スキルが何をするかと、いつ使うか。`description` と `when_to_use` を合わせたテキストはスキル一覧で 1,536 文字で切り詰められる（`skillListingMaxDescChars` で変更可能）。自動呼び出しのマッチングに重要。 |
+| `when_to_use` | Claude がスキルを呼び出すべきタイミングに関する補足情報。スキル一覧では `description` の後ろに連結され、1,536 文字の上限に算入される。 |
 | `argument-hint` | `/` の自動補完メニューに表示されるヒント（例: `"[filename] [format]"`）。 |
 | `disable-model-invocation` | `true` = ユーザーのみが `/name` で呼び出せる。Claude は自動呼び出ししない。 |
 | `user-invocable` | `false` = `/` メニューから非表示。Claude のみが自動呼び出し可能。 |
@@ -617,7 +618,7 @@ Can you help me review this code for security issues?

 ### スキルの更新

-`SKILL.md` を直接編集する。変更は次回の Claude Code 起動時に反映される。
+`SKILL.md` を直接編集し、`/reload-skills`（v2.1.152 以降）でスキルディレクトリを再スキャンする。再起動でも反映されるが必須ではない。`--add-dir` のスキルはライブ検出され、`SessionStart` フックが `reloadSkills: true` を返しても同じ再スキャンが走る。

 ```bash
 # 個人スキル
@@ -823,8 +824,8 @@ chmod +x ~/.claude/skills/my-skill/scripts/*.py
 - [フックガイド](../06-hooks/) - イベント駆動の自動化

 ---
-**最終更新**: 2026 年 4 月 24 日
-**Claude Code バージョン**: 2.1.119
+**最終更新**: 2026 年 8 月 25 日
+**Claude Code バージョン**: 2.1.245
 **情報源**:
 - https://code.claude.com/docs/en/skills
 - https://code.claude.com/docs/en/settings
@@ -125,7 +125,7 @@ hooks:  # オプション — コンポーネントスコープのフック
 | `tools` | いいえ | 特定ツールのカンマ区切りリスト。省略するとすべてのツールを継承。`Agent(agent_name)` 構文で生成可能なサブエージェントを制限できる |
 | `disallowedTools` | いいえ | サブエージェントが使用してはならないツールのカンマ区切りリスト |
 | `model` | いいえ | 使用モデル：`sonnet`、`opus`、`haiku`、フルモデル ID、または `inherit`。デフォルトは設定済みのサブエージェントモデル |
-| `permissionMode` | いいえ | `default`、`acceptEdits`、`dontAsk`、`bypassPermissions`、`plan` |
+| `permissionMode` | いいえ | `manual`（v2.1.200 で `default` から改名 — 旧名の `default` も引き続き有効）、`acceptEdits`、`dontAsk`、`bypassPermissions`、`plan`、`auto` |
 | `maxTurns` | いいえ | サブエージェントが取れるエージェント的ターンの最大数 |
 | `skills` | いいえ | プリロードするスキルのカンマ区切りリスト。起動時にスキルの全内容をサブエージェントのコンテキストに注入する |
 | `mcpServers` | いいえ | サブエージェントから利用可能にする MCP サーバー |
@@ -256,10 +256,10 @@ Claude Code には常に利用可能な組み込みサブエージェントが
 |-------------|-------|------|
 | **general-purpose** | 継承 | 複雑な多段タスク |
 | **Plan** | 継承 | プランモードのリサーチ |
-| **Explore** | Haiku | 読み取り専用のコードベース探索（quick/medium/very thorough） |
-| **Bash** | 継承 | 別コンテキストでのターミナルコマンド |
-| **statusline-setup** | Sonnet | ステータスラインの設定 |
-| **Claude Code Guide** | Haiku | Claude Code の機能に関する質問への回答 |
+| **Explore** | 継承（上限は Opus） | 読み取り専用のコードベース探索（quick/medium/very thorough） |
+| **claude** | 継承 | より専門的なエージェントに当てはまらないタスク向けの汎用エージェント。サブエージェントが利用できるすべてのツールを持つ。ディスパッチされたバックグラウンドセッションのデフォルトエージェントでもある |
+| **statusline-setup** | Sonnet | `/statusline` でステータスラインを設定するときに実行される |
+| **claude-code-guide** | Haiku | Claude Code の機能に関する質問への回答 |

 ### General-Purpose サブエージェント

@@ -298,15 +298,15 @@ Claude Code には常に利用可能な組み込みサブエージェントが
 - **"medium"** — 適度な探索。速度と徹底度のバランスが取れたデフォルトのアプローチ
 - **"very thorough"** — 複数の場所と命名規約をまたぐ網羅的分析。時間がかかる場合がある

-### Bash サブエージェント
+### claude サブエージェント

 | プロパティ | 値 |
 |-----------|-----|
 | **モデル** | 親から継承 |
-| **ツール** | Bash |
-| **目的** | 別のコンテキストウィンドウでターミナルコマンドを実行 |
+| **ツール** | サブエージェントが利用できるすべてのツール |
+| **目的** | より専門的なエージェントに当てはまらないタスク向けの汎用エージェント |

-**使用場面**：隔離されたコンテキストの恩恵を受けるシェルコマンドを実行するとき。
+**使用場面**：より専門的な組み込みエージェントに当てはまらないタスクのとき。ディスパッチされたバックグラウンドセッションのデフォルトエージェントでもあり、開始時の権限モードはそのセッションの起動方法によって決まる。

 ### Statusline Setup サブエージェント

@@ -318,7 +318,7 @@ Claude Code には常に利用可能な組み込みサブエージェントが

 **使用場面**：ステータスラインのセットアップやカスタマイズ時。

-### Claude Code Guide サブエージェント
+### Claude Code Guide サブエージェント（`claude-code-guide`）

 | プロパティ | 値 |
 |-----------|-----|
@@ -1275,8 +1275,8 @@ graph TD

 ---

-**最終更新**：2026 年 4 月 24 日
-**Claude Code バージョン**：2.1.119
+**最終更新**：2026 年 8 月 25 日
+**Claude Code バージョン**：2.1.245
 **ソース**：

 - https://code.claude.com/docs/en/sub-agents
@@ -109,6 +109,28 @@ Server-Sent Events トランスポートは `http` の登場により非推奨
 claude mcp add --transport sse legacy-server https://example.com/sse
 ```

+### WebSocket トランスポート（`ws`）
+
+WebSocket サーバーは持続的な双方向接続を保つため、Claude に対して自発的にイベントを push するリモート MCP サーバーに向いている。サーバーがリクエストに応答するだけなら HTTP を使うこと。HTTP は OAuth と `claude mcp add --transport` フラグの両方をサポートするが、WebSocket はどちらもサポートしない。
+
+`--transport` は `ws` を受け付けないため、`.mcp.json` または `claude mcp add-json` で設定する。
+
+```json
+{
+  "type": "ws",
+  "url": "wss://mcp.example.com/socket",
+  "headers": {
+    "Authorization": "Bearer YOUR_TOKEN"
+  }
+}
+```
+
+`type: "ws"` のエントリは `http` と同じ `url`、`headers`、`headersHelper`、`timeout`、`alwaysLoad` フィールドを受け付ける。認証は**ヘッダーのみ**で、WebSocket サーバー向けの OAuth フローは存在しない。
+
+> **注意**: WebSocket サーバーは `claude mcp list` の出力に表示されない。確認するには `claude mcp get <名前>` または `/mcp` パネルを使うこと。
+
+HTTP や SSE と同様に、WebSocket 接続のアイドル時間は 5 分。stdio と WebSocket にはリクエスト単位のタイマーはない。`type` のない `url` エントリはエラーとなり、有効な値として `"http"`、`"sse"`、`"ws"` が示される。
+
 ### Windows 固有の注意点

 ネイティブ Windows（WSL ではない）では、npx コマンドに `cmd /c` を使う。
@@ -1126,8 +1148,8 @@ export GITHUB_TOKEN="your_token"

 ---

-**最終更新：** 2026 年 4 月 24 日
-**Claude Code バージョン：** 2.1.119
+**最終更新：** 2026 年 8 月 25 日
+**Claude Code バージョン：** 2.1.245
 **情報源：**
 - https://code.claude.com/docs/en/mcp
 - https://code.claude.com/docs/en/changelog
@@ -61,8 +61,14 @@
 | `hooks` | フック定義の配列 | `[{ "type": "command", ... }]` |
 | `type` | フックタイプ：`"command"`（bash）、`"prompt"`（LLM）、`"http"`（webhook）、`"mcp_tool"`（MCP ツール呼び出し、v2.1.118 以降）、`"agent"`（サブエージェント） | `"command"` |
 | `command` | 実行するシェルコマンド | `"$CLAUDE_PROJECT_DIR/.claude/hooks/format.sh"` |
-| `timeout` | オプションのタイムアウト（秒、デフォルト 60） | `30` |
+| `timeout` | オプションのタイムアウト（秒）。デフォルトは command/http/mcp_tool が 600、prompt が 30、agent が 60。 | `30` |
 | `once` | `true` の場合、フックはセッションごとに 1 回のみ実行 | `true` |
+| `async` | `true` の場合、ブロックせずにバックグラウンドで実行 | `true` |
+| `asyncRewake` | `true` の場合、バックグラウンドで実行し、終了コード 2 で Claude を起こす。`async` を含意する。 | `true` |
+| `shell` | `"bash"` または `"powershell"` を受け付ける。デフォルトは `"bash"`、Git Bash が未インストールの Windows では `"powershell"`。 | `"bash"` |
+| `statusMessage` | フック実行中に表示されるカスタムスピナーメッセージ | `"フォーマット中…"` |
+
+> **注意**: 一部のイベントはデフォルトのタイムアウトを引き下げる。`UserPromptSubmit` は command / http / mcp_tool のデフォルトを 30 秒に、`MessageDisplay` は 10 秒に下げる。`SessionEnd` フックは 1.5 秒の予算を共有し、設定でより長い `timeout` を指定した場合は Claude Code が最大 60 秒まで予算を引き上げる。

 ### マッチャーパターン

@@ -171,6 +177,8 @@ LLM はプロンプトを評価し、構造化された判定を返す（詳細
 }
 ```

+> **注意**: Agent フックは実験的機能であり、変更される可能性がある。
+
 **主要プロパティ：**
 - `"type": "agent"` -- agent フックであることを示す
 - `"prompt"` -- サブエージェントへのタスク説明
@@ -1314,7 +1322,7 @@ echo $?

 | 項目 | 動作 |
 |------|------|
-| **タイムアウト** | デフォルト 60 秒、コマンドごとに設定可能 |
+| **タイムアウト** | command/http/mcp_tool はデフォルト 600 秒（prompt は 30 秒、agent は 60 秒）、フックごとに設定可能 |
 | **並列化** | マッチしたフックはすべて並列実行 |
 | **重複排除** | 同一のフックコマンドは重複排除される |
 | **環境** | カレントディレクトリで Claude Code の環境のもと実行 |
@@ -1371,8 +1379,8 @@ chmod +x ~/.claude/hooks/*.sh

 ---

-**最終更新：** 2026 年 8 月 15 日
-**Claude Code バージョン：** 2.1.233
+**最終更新：** 2026 年 8 月 25 日
+**Claude Code バージョン：** 2.1.245
 **情報源：**
 - https://code.claude.com/docs/en/hooks
 - https://code.claude.com/docs/en/changelog
@@ -49,13 +49,14 @@

 ## 巻き戻しオプション

-巻き戻しを実行すると、5 つのオプションメニューが表示される：
+巻き戻しを実行すると、6 つのオプションメニューが表示される：

 1. **Restore code and conversation** -- 指定したチェックポイントの状態にファイルとメッセージの両方を戻す
 2. **Restore conversation** -- メッセージのみ巻き戻し、現在のコードはそのまま残す
 3. **Restore code** -- ファイル変更のみ戻し、会話履歴は完全に維持する
 4. **Summarize from here** -- そこ以降の会話を AI 生成のサマリに圧縮し、コンテキストウィンドウを解放する。選択点より前のメッセージはそのまま残る。ディスク上のファイルは変更されない。元のメッセージはセッショントランスクリプトに保存される。サマリを特定のトピックに絞るための指示を任意で与えられる。
-5. **Never mind** -- 操作をキャンセルし、現在の状態に戻る
+5. **Summarize up to here** -- 逆方向の対応物。選択点より *前* のすべてを AI 生成のサマリに圧縮し、そこ以降のメッセージはそのまま残す。「Summarize from here」と合わせて、コンテキストウィンドウを双方向かつ狙った範囲で圧縮できる。ディスク上のファイルは変更されず、元のメッセージはセッショントランスクリプトに保存される。
+6. **Never mind** -- 操作をキャンセルし、現在の状態に戻る

 > **注意**: 会話を復元または要約した後、選択したメッセージの元のプロンプトが入力欄に復元され、再送信や編集ができる状態になる。

@@ -293,7 +294,7 @@ Claude Code がスナップショットを保持するのは **直近 100 件の
 1. **普通に作業する** - Claude Code が自動でチェックポイントを作成
 2. **戻りたくなったら** - `Esc` を 2 回押すか `/rewind` を使う
 3. **チェックポイントを選ぶ** - リストから巻き戻し先を選択
-4. **復元対象を選ぶ** - コードと会話、会話のみ、コードのみ、要約、キャンセルから選択
+4. **復元対象を選ぶ** - コードと会話、会話のみ、コードのみ、ここから要約、ここまで要約、キャンセルから選択
 5. **作業を続ける** - 選んだ時点に戻った状態で再開

 ### キーボードショートカット
@@ -335,8 +336,8 @@ Claude Code がスナップショットを保持するのは **直近 100 件の

 ---

-**最終更新**: 2026 年 4 月 24 日
-**Claude Code バージョン**: 2.1.119
+**最終更新**: 2026 年 8 月 25 日
+**Claude Code バージョン**: 2.1.245
 **出典**:
 - https://code.claude.com/docs/en/checkpointing
 - https://code.claude.com/docs/en/settings
@@ -612,18 +612,15 @@ Claude: [Shows linter output from bg-5002]

 ### 設定

-```json
-{
-  "backgroundTasks": {
-    "enabled": true,
-    "maxConcurrentTasks": 5,
-    "notifyOnCompletion": true,
-    "autoCleanup": true,
-    "logOutput": true
-  }
-}
+`settings.json` にバックグラウンドタスク用のブロックは存在しない。この機能は環境変数で制御する：
+
+```bash
+# バックグラウンドタスクを完全に無効化する
+export CLAUDE_CODE_DISABLE_BACKGROUND_TASKS=true
 ```

+同時実行数もバックグラウンドタスクの設定ではない。同時に実行されるエージェント数は `CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS`（デフォルトは `20`）で決まる。
+
 ---

 ## Monitor ツール（イベント駆動ストリーム）
@@ -1043,7 +1040,7 @@ Claude Code は効率向上のためキーボードショートカットをサ
 | `Ctrl+C` | 入力／生成のキャンセル |
 | `Ctrl+D` | Claude Code を終了 |
 | `Ctrl+G` | 計画を外部エディタで編集 |
-| `Ctrl+L` | ターミナル画面のクリア |
+| `Ctrl+L` | 画面の再描画（repaint のみ — 2 回押しでの `/clear` は v2.1.238 で廃止） |
 | `Ctrl+O` | 詳細出力の切り替え（推論を表示） |
 | `Ctrl+R` | 履歴の逆方向検索 |
 | `Ctrl+T` | タスクリストビューの切り替え |
@@ -1587,7 +1584,7 @@ Claude Code デスクトップアプリは、視覚的な diff レビュー、

 ### デスクトップでの権限モード

-デスクトップアプリは CLI と同じ 4 つの権限モードをサポートする：
+デスクトップアプリは CLI と同じ権限モードをサポートする：

 | モード | 挙動 |
 |------|----------|
@@ -2111,8 +2108,8 @@ Claude Code および関連機能の詳細：

 ---

-**最終更新：** 2026 年 8 月 15 日
-**Claude Code バージョン：** 2.1.233
+**最終更新：** 2026 年 8 月 25 日
+**Claude Code バージョン：** 2.1.245
 **情報源：**
 - https://code.claude.com/docs/en/permission-modes
 - https://code.claude.com/docs/en/commands
@@ -92,7 +92,7 @@
 | `/stickers` | セッションステッカーを表示 | 楽しいご褒美 |
 | `/fast` | 高速出力モード切替 | 応答の高速化 |
 | `/terminal-setup` | ターミナル統合を設定 | ターミナル機能のセットアップ |
-| `/undo` | `/rewind` のエイリアス（v2.1.108） | `/rewind` と同じ |
+| `/undo` | **公式ドキュメントに記載なし** — v2.1.108 で `/rewind` のエイリアスとして追加されたが、公式のコマンドリファレンスには一切登場しない | 代わりに `/rewind`（または `Esc Esc`）を使う |
 | `/upgrade` | アップデート確認 | バージョン管理 |
 | `/team-onboarding` | このプロジェクトの Claude Code 利用状況からオンボーディングガイドを生成 | 新メンバーのオンボーディング（v2.1.101） |
 | `/code-review ultra` | 現在の変更に対するクラウドのマルチエージェントコードレビューを実行。`/ultrareview` はエイリアスとして残っており、`/code-review ultra` が推奨の呼び出し。Pro と Max では 3 回まで無料、以降は使用クレジットが必要 | マージ前の深いマルチエージェントレビュー（v2.1.112） |
@@ -151,10 +151,10 @@ Claude Code はツール使用の許可を制御する 6 つの権限モード
 |-------|-------------|-------|-------|-------------|
 | **general-purpose** | 多段タスク、調査 | 全ツール | モデル継承 | 複雑な調査、複数ファイルタスク |
 | **Plan** | 実装計画 | Read、Glob、Grep、Bash | モデル継承 | アーキテクチャ設計、計画 |
-| **Explore** | コードベース探索 | Read、Glob、Grep | Haiku 4.5 | 高速検索、コード理解 |
-| **Bash** | コマンド実行 | Bash | モデル継承 | Git 操作、ターミナル作業 |
+| **Explore** | コードベース探索 | Read、Glob、Grep | 継承（上限は Opus） | 高速検索、コード理解 |
+| **claude** | より専門的なエージェントに当てはまらないタスク向けの汎用エージェント | 全ツール | モデル継承 | 専用エージェントがないタスク、ディスパッチされたバックグラウンドセッションのデフォルト |
 | **statusline-setup** | ステータスライン設定 | Bash、Read、Write | Sonnet 4.6 | ステータスライン表示の設定 |
-| **Claude Code Guide** | ヘルプとドキュメント | Read、Glob、Grep | Haiku 4.5 | ヘルプ、機能学習 |
+| **claude-code-guide** | ヘルプとドキュメント | Read、Glob、Grep | Haiku 4.5 | ヘルプ、機能学習 |

 ### サブエージェント設定フィールド

@@ -449,7 +449,7 @@ cp 02-memory/personal-CLAUDE.md ~/.claude/CLAUDE.md
 | **/proactive** | `/loop` のエイリアス — 同じ反復タスク動作（v2.1.105） | `/proactive` を `/loop` と互換に利用 |
 | **/recap** | 既存セッションに戻ったときに要約を表示（v2.1.108） | 離席後に `/recap` を実行して文脈を取り戻す |
 | **/tui** | フルスクリーン TUI モードを切替してちらつきのない描画（v2.1.110） | フルスクリーンターミナルや tmux で `/tui` を使う |
-| **/undo** | `/rewind` のエイリアス — 直前のチェックポイントへ戻す（v2.1.108） | `/undo` を `/rewind` と互換に利用 |
+| **/undo** | **公式ドキュメントに記載なし** — v2.1.108 で `/rewind` のエイリアスとして追加されたが、公式のコマンドリファレンスには一切登場しない | 代わりに `/rewind`（または `Esc Esc`）を使う |
 | **Monitor ツール** | バックグラウンドコマンドの stdout を監視し、ポーリングではなくイベントで反応（v2.1.98+） | [高度な機能](09-advanced-features/) 経由で Monitor ツールを利用 |
 | **/team-onboarding** | プロジェクトの Claude Code 設定からオンボーディングガイドを自動生成（v2.1.101） | プロジェクトで `/team-onboarding` を実行 |
 | **リモートコントロール** | API 経由で Claude Code セッションを遠隔制御 | リモートコントロール API でプロンプト送信と応答取得をプログラム的に行う |
@@ -532,8 +532,8 @@ chmod +x ~/.claude/hooks/*.sh

 ---

-**最終更新**：2026 年 8 月 15 日
-**Claude Code バージョン**：2.1.233
+**最終更新**：2026 年 8 月 25 日
+**Claude Code バージョン**：2.1.245
 **情報源**：
 - https://code.claude.com/docs/en/overview
 - https://code.claude.com/docs/en/commands
@@ -877,8 +877,8 @@ MIT License — [LICENSE](../LICENSE) を参照。利用、変更、再配布は

 ---

-**最終更新**：2026 年 8 月 15 日
-**Claude Code バージョン**：2.1.233
+**最終更新**：2026 年 8 月 25 日
+**Claude Code バージョン**：2.1.245
 **情報源**：
 - https://code.claude.com/docs/en/hooks
 - https://code.claude.com/docs/en/overview
@@ -15,7 +15,6 @@ Claude How To プロジェクトのセキュリティを私たちは重視して
 | バージョン | ステータス | サポート期限 |
 |---------|--------|---------------|
 | Latest（main） | ✅ アクティブ | 現在 + 6 か月 |
-| 1.x リリース | ✅ アクティブ | 次のメジャーバージョンまで |

 **注：** 教育用ガイドプロジェクトとして、私たちは従来のバージョンサポートよりも、現在のベストプラクティスとドキュメントセキュリティの維持に重点を置いている。更新は main ブランチに直接適用される。

@@ -330,7 +329,10 @@ pip-audit
 - プロジェクトの範囲が変わったとき
 - 最低でも年 1 回

-**Last Updated**: January 2026
+**最終更新**：2026 年 8 月 25 日
+**Claude Code バージョン**：2.1.245
+**情報源**：
+- https://code.claude.com/docs/en/overview
 **Next Review**: January 2027

 ---
@@ -603,12 +603,13 @@ docs(skills): Add comprehensive code review skill

 ```markdown
 ---
-**Last Updated**: March 2026
-**Claude Code Version**: 2.1.97
+**Last Updated**: August 2026
+**Claude Code Version**: 2.1.245
 **Compatible Models**: Claude Sonnet 4.6, Claude Opus 4.7, Claude Haiku 4.5
 ```

- 月＋年の形式を使う（例：「March 2026」）
+- 月＋年の形式を使う（例：「August 2026」）
+- バージョンは、ここに示した値ではなく、現在の同期実行で取得したものを使う
 - 機能変更時にバージョンを更新する
 - 互換モデルをすべてリストする

@@ -637,10 +638,10 @@ docs(skills): Add comprehensive code review skill

 ---

-**Last Updated**: April 24, 2026
-**Claude Code Version**: 2.1.119
-**Sources**:
+**最終更新**：2026 年 8 月 25 日
+**Claude Code バージョン**：2.1.245
+**情報源**：
 - https://code.claude.com/docs/en/overview
 - https://code.claude.com/docs/en/changelog
 - https://www.anthropic.com/news/claude-opus-4-7
-**Compatible Models**: Claude Sonnet 4.6, Claude Opus 4.7, Claude Haiku 4.5
+**互換モデル**：Claude Sonnet 4.6、Claude Opus 4.7、Claude Haiku 4.5
@@ -3108,16 +3108,14 @@ User: \
    "enabled": true,
    "showThinkingProcess": true
  },
-  "backgroundTasks": {
-    "enabled": true,
-    "maxConcurrentTasks": 5
-  },
  "permissions": {
    "defaultMode": "default"
  }
 }
 ```

+`settings.json` にバックグラウンドタスク用のブロックは存在しない。この機能は環境変数 `CLAUDE_CODE_DISABLE_BACKGROUND_TASKS` で制御し、同時実行数は `CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS`（デフォルトは `20`）で決まる。
+
 **詳細ガイド：** [09-advanced-features/](09-advanced-features/)

 ---
@@ -114,7 +114,7 @@
 ### Останні зміни

 - `/fork` перейменовано на `/branch`, `/fork` залишено як аліас (v2.1.77)
- `/output-style` застаріла (v2.1.73) і видалена (v2.1.91) — стилі виводу все ще доступні через `/config` → Output style або налаштування `outputStyle`; вбудовані: Default, Proactive, Explanatory, Learning
+- `/output-style` застаріла (v2.1.73) і видалена (v2.1.91) — стилі виводу все ще доступні через `/config` → Output style або налаштування `outputStyle`; вбудовані: Default, Proactive, Explanatory, Learning, Concise (додано у v2.1.237)
 - `/review` став повним псевдонімом `/code-review` — ті самі цілі, рівні зусиль і прапорці (v2.1.223). Історія: спершу він перейшов на рушій `/code-review medium` у v2.1.186, залишаючись доступним лише для PR
 - Додано команду `/effort` з рівнем `max` для Opus 4.6
 - Додано команду `/voice` для голосового вводу push-to-talk
@@ -637,7 +637,9 @@ Claude завантажить CLAUDE.md із зазначеного додатк
 ````markdown
 # API Module Standards

-This file overrides root CLAUDE.md for everything in /src/api/
+This file supplements root CLAUDE.md for everything in /src/api/. Memory files are
+concatenated, not overridden — the root CLAUDE.md still applies, and Claude Code
+loads this file on demand when it reads files in this subtree.

 ## API-Specific Standards

@@ -1079,7 +1081,9 @@ Claude запропонує обрати, який файл пам'яті оно
   cat > /path/to/directory/CLAUDE.md << 'EOF'
   # [Directory Name] Standards

-   This file overrides root CLAUDE.md for this directory.
+   This file supplements root CLAUDE.md for this directory. Memory files are
+   concatenated, not overridden — Claude Code loads this file on demand when it
+   reads files in this directory.

   ## [Specific Standards]
   EOF
@@ -1154,6 +1158,8 @@ Claude запропонує обрати, який файл пам'яті оно
 - [Офіційна документація пам'яті](https://code.claude.com/docs/en/memory) — документація Anthropic

 ---
-**Останнє оновлення**: 9 квітня 2026
-**Версія Claude Code**: 2.1.97
+**Останнє оновлення**: 25 серпня 2026
+**Версія Claude Code**: 2.1.245
+**Джерела**:
+- https://code.claude.com/docs/en/memory
 **Сумісні моделі**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -96,7 +96,7 @@ sequenceDiagram
 | **Project** | `.claude/skills/<skill-name>/SKILL.md` | Команда | Так (через git) | Командні стандарти |
 | **Plugin** | `<plugin>/skills/<skill-name>/SKILL.md` | Де увімкнено | Залежить | В складі плагінів |

-Коли навички мають однакову назву на різних рівнях, вищі пріоритетні розташування перемагають: **enterprise > project > personal**. Навички плагінів використовують простір імен `plugin-name:skill-name`, тому конфлікти неможливі.
+Коли навички мають однакову назву на різних рівнях, вищі пріоритетні розташування перемагають: **enterprise > personal > project**. Навички плагінів використовують простір імен `plugin-name:skill-name`, тому конфлікти неможливі.

 ### Автоматичне виявлення

@@ -140,7 +140,7 @@ Show concrete examples of using this Skill.
 ### Обов'язкові поля

 - **name**: тільки малі літери, цифри, дефіси (макс. 64 символи). Не може містити "anthropic" або "claude".
- **description**: що навичка робить І коли її використовувати (макс. 1024 символи). Це критично для того, щоб Claude знав, коли активувати навичку.
+- **description**: що навичка робить І коли її використовувати. Об'єднаний текст `description` + `when_to_use` обрізається до **1 536 символів** у переліку навичок (налаштовується через `skillListingMaxDescChars`). Це критично для того, щоб Claude знав, коли активувати навичку.

 ### Додаткові поля фронтматера

@@ -170,7 +170,8 @@ paths: "src/api/**/*.ts"               # Glob-патерни, що обмежу
 | Поле | Опис |
 |------|------|
 | `name` | Тільки малі літери, цифри, дефіси (макс. 64 символи). Не може містити "anthropic" або "claude". |
-| `description` | Що навичка робить І коли її використовувати (макс. 1024 символи). Критично для автоматичного зіставлення. |
+| `description` | Що навичка робить І коли її використовувати. Об'єднаний текст `description` + `when_to_use` обрізається до 1 536 символів у переліку навичок (налаштовується через `skillListingMaxDescChars`). Критично для автоматичного зіставлення. |
+| `when_to_use` | Додатковий контекст про те, коли Claude має викликати навичку. Додається до `description` у переліку навичок і враховується в ліміті 1 536 символів. |
 | `argument-hint` | Підказка в меню автодоповнення `/` (наприклад, `"[filename] [format]"`). |
 | `disable-model-invocation` | `true` = тільки користувач може викликати через `/name`. Claude ніколи не викличе автоматично. |
 | `user-invocable` | `false` = приховано з меню `/`. Тільки Claude може викликати автоматично. |
@@ -617,7 +618,7 @@ Can you help me review this code for security issues?

 ### Оновлення навички

-Редагуйте файл `SKILL.md` безпосередньо. Зміни набувають чинності при наступному запуску Claude Code.
+Редагуйте файл `SKILL.md` безпосередньо, а потім виконайте `/reload-skills` (v2.1.152+), щоб повторно просканувати каталоги навичок. Перезапуск Claude Code також працює, але не обов'язковий — навички з каталогів `--add-dir` підхоплюються наживо, а хук `SessionStart`, що повертає `reloadSkills: true`, запускає те саме пересканування.

 ```bash
 # Персональна навичка
@@ -810,6 +811,8 @@ chmod +x ~/.claude/skills/my-skill/scripts/*.py
 - [Посібник хуків](../06-hooks/) — автоматизація за подіями

 ---
-**Останнє оновлення**: 9 квітня 2026
-**Версія Claude Code**: 2.1.97
+**Останнє оновлення**: 25 серпня 2026
+**Версія Claude Code**: 2.1.245
+**Джерела**:
+- https://code.claude.com/docs/en/skills
 **Сумісні моделі**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -126,7 +126,7 @@ to solving problems.
 | `tools` | Ні | Список інструментів через кому. Пропустіть для успадкування всіх. Підтримує синтаксис `Agent(agent_name)` для обмеження створюваних субагентів |
 | `disallowedTools` | Ні | Список заборонених інструментів через кому |
 | `model` | Ні | Модель: `sonnet`, `opus`, `haiku`, повний ID моделі або `inherit`. За замовч. — налаштована модель субагента |
-| `permissionMode` | Ні | `default`, `acceptEdits`, `dontAsk`, `bypassPermissions`, `plan` |
+| `permissionMode` | Ні | `manual` (перейменовано з `default` у v2.1.200 — стара назва `default` досі приймається), `acceptEdits`, `dontAsk`, `bypassPermissions`, `plan`, `auto` |
 | `maxTurns` | Ні | Максимальна кількість агентних кроків |
 | `skills` | Ні | Список навичок через кому для попереднього завантаження. Вставляє повний вміст навички в контекст при запуску |
 | `mcpServers` | Ні | MCP-сервери, доступні субагенту |
@@ -214,10 +214,10 @@ Claude Code включає кілька вбудованих субагенті
 |-------|--------|-------------|
 | **general-purpose** | Успадковує | Складні, багатокрокові завдання |
 | **Plan** | Успадковує | Дослідження для режиму планування |
-| **Explore** | Haiku | Дослідження кодової бази лише для читання (швидке/середнє/дуже ретельне) |
-| **Bash** | Успадковує | Термінальні команди в окремому контексті |
-| **statusline-setup** | Sonnet | Налаштування рядка стану |
-| **Claude Code Guide** | Haiku | Відповіді на питання про функції Claude Code |
+| **Explore** | Успадковує (максимум Opus) | Дослідження кодової бази лише для читання (швидке/середнє/дуже ретельне) |
+| **claude** | Успадковує | Універсальний агент для задач, які не підходять спеціалізованим агентам; має всі інструменти, доступні субагентам. Також типовий агент для диспетчеризованої фонової сесії |
+| **statusline-setup** | Sonnet | Запускається, коли ви використовуєте `/statusline` для налаштування рядка стану |
+| **claude-code-guide** | Haiku | Відповіді на питання про функції Claude Code |

 ### Субагент General-Purpose

@@ -255,15 +255,15 @@ Claude Code включає кілька вбудованих субагенті
 - **"medium"** — помірне дослідження, баланс швидкості та ретельності, стандартний підхід
 - **"very thorough"** — комплексний аналіз у кількох розташуваннях та конвенціях іменування, може зайняти більше часу

-### Субагент Bash
+### Субагент claude

 | Властивість | Значення |
 |-------------|----------|
 | **Модель** | Успадковує від батька |
-| **Інструменти** | Bash |
-| **Призначення** | Виконання термінальних команд в окремому контекстному вікні |
+| **Інструменти** | Усі інструменти, доступні субагентам |
+| **Призначення** | Універсальний агент для задач, які не підходять спеціалізованим агентам |

-**Коли використовується**: Виконання команд оболонки, які виграють від ізольованого контексту.
+**Коли використовується**: Коли задача не відповідає жодному спеціалізованішому вбудованому агенту. Це також типовий агент для диспетчеризованої фонової сесії; режим дозволів на старті залежить від того, як цю сесію було запущено.

 ### Субагент Statusline Setup

@@ -275,7 +275,7 @@ Claude Code включає кілька вбудованих субагенті

 **Коли використовується**: Налаштування або кастомізація рядка стану.

-### Субагент Claude Code Guide
+### Субагент Claude Code Guide (`claude-code-guide`)

 | Властивість | Значення |
 |-------------|----------|
@@ -1140,6 +1140,8 @@ graph TD
 - [Посібник хуків](../06-hooks/) — для автоматизації за подіями

 ---
-**Останнє оновлення**: 9 квітня 2026
-**Версія Claude Code**: 2.1.97
+**Останнє оновлення**: 25 серпня 2026
+**Версія Claude Code**: 2.1.245
+**Джерела**:
+- https://code.claude.com/docs/en/sub-agents
 **Сумісні моделі**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -109,6 +109,28 @@ claude mcp add --transport stdio myserver --env KEY=value -- npx server
 claude mcp add --transport sse legacy-server https://example.com/sse
 ```

+### WebSocket-транспорт (`ws`)
+
+WebSocket-сервери тримають постійне двонаправлене зʼєднання, що підходить віддаленим MCP-серверам, які надсилають події до Claude без запиту. Використовуйте HTTP, якщо ваш сервер лише відповідає на запити, оскільки HTTP підтримує OAuth і прапорець `claude mcp add --transport`, а WebSocket — жодного з них.
+
+Оскільки `--transport` не приймає `ws`, налаштуйте його в `.mcp.json` або через `claude mcp add-json`:
+
+```json
+{
+  "type": "ws",
+  "url": "wss://mcp.example.com/socket",
+  "headers": {
+    "Authorization": "Bearer YOUR_TOKEN"
+  }
+}
+```
+
+Запис `type: "ws"` приймає ті самі поля `url`, `headers`, `headersHelper`, `timeout` і `alwaysLoad`, що й `http`. Автентифікація **лише через заголовки** — потоку OAuth для WebSocket-серверів немає.
+
+> **Примітка**: WebSocket-сервери не зʼявляються у виводі `claude mcp list`. Для перевірки використовуйте `claude mcp get <назва>` або панель `/mcp`.
+
+Як і HTTP та SSE, WebSocket-зʼєднання використовує 5-хвилинне вікно простою; stdio і WebSocket не мають таймера на окремий запит. Запис `url` без `type` призводить до помилки, яка називає `"http"`, `"sse"`, `"ws"` як допустимі значення.
+
 ### Примітка для Windows

 На нативній Windows (не WSL) використовуйте `cmd /c` для команд npx:
@@ -1097,6 +1119,8 @@ export GITHUB_TOKEN="your_token"
 - [Документація Claude API](https://docs.anthropic.com)

 ---
-**Останнє оновлення**: 9 квітня 2026
-**Версія Claude Code**: 2.1.97
+**Останнє оновлення**: 25 серпня 2026
+**Версія Claude Code**: 2.1.245
+**Джерела**:
+- https://code.claude.com/docs/en/mcp
 **Сумісні моделі**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -59,10 +59,16 @@
 |------|------|---------|
 | `matcher` | Шаблон для відповідності назвам інструментів (чутливий до регістру) | `"Write"`, `"Edit\|Write"`, `"*"` |
 | `hooks` | Масив визначень хуків | `[{ "type": "command", ... }]` |
-| `type` | Тип хука: `"command"` (bash), `"prompt"` (LLM), `"http"` (вебхук) або `"agent"` (субагент) | `"command"` |
+| `type` | Тип хука: `"command"` (bash), `"prompt"` (LLM), `"http"` (вебхук), `"mcp_tool"` (виклик MCP-інструмента, v2.1.118+) або `"agent"` (субагент) | `"command"` |
 | `command` | Shell-команда для виконання | `"$CLAUDE_PROJECT_DIR/.claude/hooks/format.sh"` |
-| `timeout` | Необовʼязковий таймаут у секундах (за замовчуванням 60) | `30` |
+| `timeout` | Необовʼязковий таймаут у секундах. За замовчуванням: 600 для command/http/mcp_tool, 30 для prompt, 60 для agent. | `30` |
 | `once` | Якщо `true`, хук запускається лише один раз за сесію | `true` |
+| `async` | Якщо `true`, виконується у фоні без блокування | `true` |
+| `asyncRewake` | Якщо `true`, виконується у фоні та будить Claude при коді виходу 2. Неявно вмикає `async`. | `true` |
+| `shell` | Приймає `"bash"` або `"powershell"`. За замовчуванням `"bash"`, або `"powershell"` у Windows, якщо Git Bash не встановлено. | `"bash"` |
+| `statusMessage` | Власне повідомлення індикатора, що показується під час роботи хука | `"Форматування…"` |
+
+> **Примітка**: Деякі події знижують типовий таймаут. `UserPromptSubmit` знижує типове значення для command, http і mcp_tool до 30 секунд, а `MessageDisplay` — до 10 секунд. Хуки `SessionEnd` мають спільний бюджет 1,5 секунди; якщо у ваших налаштуваннях указано довший `timeout` для окремого хука, Claude Code піднімає бюджет відповідно, до 60 секунд.

 ### Шаблони matcher

@@ -169,6 +175,8 @@ LLM оцінює промпт і повертає структуроване р
 }
 ```

+> **Примітка**: Agent-хуки є експериментальними і можуть змінитися.
+
 **Ключові властивості:**
 - `"type": "agent"` — ідентифікує як agent-хук
 - `"prompt"` — опис завдання для субагента
@@ -529,6 +537,8 @@ hooks:
 }
 ```

+> **`retry` (PermissionDenied)**: Використайте JSON `hookSpecificOutput.retry: true`, щоб повідомити модель, що вона може повторити відхилений виклик інструмента.
+
 ## Змінні оточення

 | Змінна | Доступність | Опис |
@@ -1205,8 +1215,8 @@ chmod +x ~/.claude/hooks/*.sh
 - **[Посібник з памʼяті](../02-memory/)** — конфігурація постійного контексту

 ---
-**Останнє оновлення**: 15 серпня 2026
-**Версія Claude Code**: 2.1.233
+**Останнє оновлення**: 25 серпня 2026
+**Версія Claude Code**: 2.1.245
 **Джерела**:
 - https://code.claude.com/docs/en/hooks
 **Сумісні моделі**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -573,18 +573,15 @@ Claude: [Shows linter output from bg-5002]

 ### Конфігурація

-```json
-{
-  "backgroundTasks": {
-    "enabled": true,
-    "maxConcurrentTasks": 5,
-    "notifyOnCompletion": true,
-    "autoCleanup": true,
-    "logOutput": true
-  }
-}
+Немає блоку `settings.json` для фонових завдань. Функція керується змінною середовища:
+
+```bash
+# Повністю вимкнути фонові завдання
+export CLAUDE_CODE_DISABLE_BACKGROUND_TASKS=true
 ```

+Паралельність теж не є налаштуванням фонових завдань — кількість одночасно запущених агентів визначає `CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS` (за замовчуванням `20`).
+
 ---

 ## Заплановані завдання
@@ -950,7 +947,7 @@ Claude Code підтримує клавіатурні скорочення дл
 | `Ctrl+C` | Скасувати поточний ввід/генерацію |
 | `Ctrl+D` | Вийти з Claude Code |
 | `Ctrl+G` | Редагувати план у зовнішньому редакторі |
-| `Ctrl+L` | Очистити екран терміналу |
+| `Ctrl+L` | Перемалювати екран (лише перемальовування — подвійне натискання для `/clear` вилучено у v2.1.238) |
 | `Ctrl+O` | Перемкнути детальний вивід (перегляд міркувань) |
 | `Ctrl+R` | Зворотний пошук в історії |
 | `Ctrl+T` | Перемкнути перегляд списку завдань |
@@ -1436,7 +1433,7 @@ claude --teleport

 ### Режими дозволів у десктопному додатку

-Десктопний додаток підтримує ті самі 4 режими дозволів, що й CLI:
+Десктопний додаток підтримує ті самі режими дозволів, що й CLI:

 | Режим | Поведінка |
 |------|----------|
@@ -1936,8 +1933,8 @@ claude --teammate-mode in-process
 - [Офіційна документація Agent Teams](https://code.claude.com/docs/en/agent-teams)

 ---
-**Останнє оновлення**: 15 серпня 2026
-**Версія Claude Code**: 2.1.233
+**Останнє оновлення**: 25 серпня 2026
+**Версія Claude Code**: 2.1.245
 **Джерела**:
 - https://code.claude.com/docs/en/commands
 **Сумісні моделі**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -144,10 +144,10 @@ Claude Code підтримує 6 режимів дозволів, що конт
 |-------|------|-------------|--------|---------------------|
 | **general-purpose** | Багатокрокові завдання, дослідження | Усі інструменти | Успадкована | Складні дослідження, багатофайлові завдання |
 | **Plan** | Планування реалізації | Read, Glob, Grep, Bash | Успадкована | Проєктування архітектури, планування |
-| **Explore** | Дослідження кодової бази | Read, Glob, Grep | Haiku 4.5 | Швидкий пошук, розуміння коду |
-| **Bash** | Виконання команд | Bash | Успадкована | Git-операції, термінальні завдання |
+| **Explore** | Дослідження кодової бази | Read, Glob, Grep | Успадковує (максимум Opus) | Швидкий пошук, розуміння коду |
+| **claude** | Універсальний агент для задач, які не підходять спеціалізованим агентам | Усі інструменти | Успадкована | Задачі без спеціалізованого агента; типовий агент для диспетчеризованої фонової сесії |
 | **statusline-setup** | Налаштування рядка стану | Bash, Read, Write | Sonnet 4.6 | Конфігурація рядка стану |
-| **Claude Code Guide** | Довідка та документація | Read, Glob, Grep | Haiku 4.5 | Отримання допомоги, вивчення функцій |
+| **claude-code-guide** | Довідка та документація | Read, Glob, Grep | Haiku 4.5 | Отримання допомоги, вивчення функцій |

 ### Поля конфігурації субагентів

@@ -659,8 +659,8 @@ Claude:

 ---

-**Останнє оновлення**: 15 серпня 2026
-**Версія Claude Code**: 2.1.233
+**Останнє оновлення**: 25 серпня 2026
+**Версія Claude Code**: 2.1.245
 **Джерела**:
 - https://code.claude.com/docs/en/hooks
 **Сумісні моделі**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -15,7 +15,6 @@
 | Версія | Статус | Підтримка до |
 |--------|--------|--------------|
 | Остання (main) | ✅ Активна | Поточна + 6 місяців |
-| Релізи 1.x | ✅ Активна | До наступної мажорної версії |

 **Примітка**: Як освітній проєкт-довідник, ми зосереджуємося на підтримці актуальних найкращих практик та безпеки документації, а не на традиційній підтримці версій. Оновлення застосовуються безпосередньо до гілки main.

@@ -330,7 +329,10 @@ pip-audit
 - При зміні обсягу проєкту
 - Щорічно як мінімум

-**Останнє оновлення**: Квітень 2026
+**Останнє оновлення**: 25 серпня 2026
+**Версія Claude Code**: 2.1.245
+**Джерела**:
+- https://code.claude.com/docs/en/overview
 **Наступний перегляд**: Квітень 2027

 ---
@@ -601,12 +601,13 @@ README уроків завершуються блоком метаданих:

 ```markdown
 ---
-**Last Updated**: March 2026
-**Claude Code Version**: 2.1.97
+**Last Updated**: August 2026
+**Claude Code Version**: 2.1.245
 **Compatible Models**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
 ```

- Використовуйте формат місяць + рік (напр., "March 2026")
+- Використовуйте формат місяць + рік (напр., "August 2026")
+- Використовуйте версію, зафіксовану поточним запуском синхронізації, а не значення, показане тут
 - Оновлюйте версію при зміні функцій
 - Перелічуйте всі сумісні моделі

@@ -635,7 +636,7 @@ README уроків завершуються блоком метаданих:

 ---

-**Останнє оновлення**: 15 серпня 2026
-**Версія Claude Code**: 2.1.233
+**Останнє оновлення**: 25 серпня 2026
+**Версія Claude Code**: 2.1.245
 **Джерела**:
 - https://code.claude.com/docs/en/overview
@@ -3104,16 +3104,14 @@ User: \
    "enabled": true,
    "showThinkingProcess": true
  },
-  "backgroundTasks": {
-    "enabled": true,
-    "maxConcurrentTasks": 5
-  },
  "permissions": {
    "defaultMode": "manual"
  }
 }
 ```

+Немає блоку `settings.json` для фонових завдань — функція керується змінною середовища `CLAUDE_CODE_DISABLE_BACKGROUND_TASKS`, а паралельність — `CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS` (за замовчуванням `20`).
+
 **Дивіться**: [09-advanced-features/](09-advanced-features/) для повного посібника

 ---
@@ -110,7 +110,7 @@ Những skills này được gửi kèm với Claude Code và được gọi nh
 ### Thay Đổi Gần Đây

 - `/fork` và `/subtask` hoán đổi vai trò trong **v2.1.212**. `/fork` giờ sao chép cuộc hội thoại thành một phiên nền độc lập mới; hành vi subagent-được-fork mà nó từng có đã chuyển sang lệnh mới `/subtask`. Lịch sử: `/fork` là bí danh của `/branch` từ v2.1.77 đến v2.1.161; từ v2.1.161 đến v2.1.211 nó khởi chạy một subagent được fork (việc mà `/subtask` làm bây giờ). Khi tắt agent view, `/subtask` không khả dụng và `/fork` giữ hành vi subagent-được-fork
- `/output-style` đã lỗi thời (v2.1.73) và bị xóa (v2.1.91) — output styles vẫn có sẵn qua `/config` → Output style hoặc setting `outputStyle`; các bản tích hợp sẵn là Default, Proactive, Explanatory, và Learning
+- `/output-style` đã lỗi thời (v2.1.73) và bị xóa (v2.1.91) — output styles vẫn có sẵn qua `/config` → Output style hoặc setting `outputStyle`; các bản tích hợp sẵn là Default, Proactive, Explanatory, Learning, và Concise (thêm ở v2.1.237)
 - `/review` trở thành bí danh đầy đủ của `/code-review` — cùng target, mức effort và flags (v2.1.223). Lịch sử: ban đầu nó chuyển sang engine `/code-review medium` trong v2.1.186 nhưng vẫn chỉ dùng cho PR
 - Lệnh `/effort` được thêm với mức `max` yêu cầu Opus 4.6
 - Lệnh `/voice` được thêm cho nhập liệu giọng nói push-to-talk
@@ -644,7 +644,9 @@ Claude sẽ tải CLAUDE.md từ thư mục bổ sung được chỉ định cù
 ```markdown
 # API Module Standards

-This file overrides root CLAUDE.md for everything in /src/api/
+This file supplements root CLAUDE.md for everything in /src/api/. Memory files are
+concatenated, not overridden — the root CLAUDE.md still applies, and Claude Code
+loads this file on demand when it reads files in this subtree.

 ## API-Specific Standards

@@ -1078,7 +1080,9 @@ Claude sẽ nhắc bạn chọn file bộ nhớ nào để cập nhật.
   cat > /path/to/directory/CLAUDE.md << 'EOF'
   # [Directory Name] Standards

-   This file overrides root CLAUDE.md for this directory.
+   This file supplements root CLAUDE.md for this directory. Memory files are
+   concatenated, not overridden — Claude Code loads this file on demand when it
+   reads files in this directory.

   ## [Specific Standards]
   EOF
@@ -1154,6 +1158,8 @@ Claude sẽ nhắc bạn chọn file bộ nhớ nào để cập nhật.

 ---

-**Cập Nhật Lần Cuối**: Tháng 4 năm 2026
-**Phiên Bản Claude Code**: 2.1+
+**Cập Nhật Lần Cuối**: Ngày 25 tháng 8 năm 2026
+**Phiên Bản Claude Code**: 2.1.245
+**Nguồn**:
+- https://code.claude.com/docs/en/memory
 **Các Mô Hình Tương Thích**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -92,7 +92,7 @@ sequenceDiagram
 | **Project** | `.claude/skills/<skill-name>/SKILL.md` | Đội | Có (qua git) | Tiêu chuẩn đội |
 | **Plugin** | `<plugin>/skills/<skill-name>/SKILL.md` | Nơi được bật | Phụ thuộc | Được gói với plugins |

-Khi skills chia sẻ cùng tên qua các cấp, vị trí ưu tiên cao hơn thắng: **enterprise > project > personal**. Skills cấp project ghi đè skills cá nhân theo mặc định; setting `skillOverrides` (v2.1.129+) điều chỉnh hành vi đó. Plugin skills sử dụng namespace `plugin-name:skill-name`, vì vậy chúng không thể xung đột.
+Khi skills chia sẻ cùng tên qua các cấp, vị trí ưu tiên cao hơn thắng: **enterprise > personal > project**. Skills cá nhân ghi đè skills cấp project theo mặc định; setting `skillOverrides` (v2.1.129+) điều chỉnh hành vi đó. Plugin skills sử dụng namespace `plugin-name:skill-name`, vì vậy chúng không thể xung đột.

 ### Khám Phá Tự Động

@@ -136,7 +136,7 @@ Hiển thị các ví dụ cụ thể về việc sử dụng Skill này.
 ### Các Trường Bắt Buộc

 - **name**: chỉ chữ thường, số, gạch ngang (tối đa 64 ký tự). Không thể chứa "anthropic" hoặc "claude".
- **description**: những gì Skill làm VÀ khi nào sử dụng nó (tối đa 1024 ký tự). Điều này quan trọng để Claude biết khi nào kích hoạt skill.
+- **description**: những gì Skill làm VÀ khi nào sử dụng nó. Nội dung `description` + `when_to_use` gộp lại bị cắt ở **1.536 ký tự** trong danh sách skill (có thể điều chỉnh qua `skillListingMaxDescChars`). Điều này quan trọng để Claude biết khi nào kích hoạt skill.

 ### Các Trường Frontmatter Tùy Chọn

@@ -165,7 +165,8 @@ hooks:                                      # Hooks theo phạm vi skill
 | Trường | Mô Tả |
 |-------|-------------|
 | `name` | Chỉ chữ thường, số, gạch ngang (tối đa 64 ký tự). Không thể chứa "anthropic" hoặc "claude". |
-| `description` | Những gì Skill làm VÀ khi nào sử dụng nó (tối đa 1024 ký tự). Quan trọng cho phù hợp kích hoạt tự động. |
+| `description` | Những gì Skill làm VÀ khi nào sử dụng nó. Nội dung `description` + `when_to_use` gộp lại bị cắt ở 1.536 ký tự trong danh sách skill (có thể điều chỉnh qua `skillListingMaxDescChars`). Quan trọng cho phù hợp kích hoạt tự động. |
+| `when_to_use` | Bối cảnh bổ sung về thời điểm Claude nên gọi skill. Được nối vào `description` trong danh sách skill và tính vào giới hạn 1.536 ký tự. |
 | `argument-hint` | Gợi ý hiển thị trong menu autocomplete `/` (ví dụ: `"[filename] [format]"`). |
 | `disable-model-invocation` | `true` = chỉ người dùng có thể gọi qua `/name`. Claude sẽ không bao giờ tự gọi. |
 | `user-invocable` | `false` = ẩn từ menu `/`. Chỉ Claude có thể gọi nó tự động. |
@@ -611,7 +612,7 @@ Can you help me review this code for security issues?

 ### Cập Nhật Một Skill

-Chỉnh sửa file `SKILL.md` trực tiếp. Các thay đổi có hiệu lực vào lần khởi động Claude Code tiếp theo.
+Chỉnh sửa file `SKILL.md` trực tiếp, sau đó chạy `/reload-skills` (v2.1.152+) để quét lại các thư mục skill. Khởi động lại Claude Code cũng được nhưng không bắt buộc — skills trong thư mục `--add-dir` được nhận diện tức thời, và hook `SessionStart` trả về `reloadSkills: true` cũng kích hoạt lần quét lại tương tự.

 ```bash
 # Personal Skill
@@ -805,6 +806,8 @@ Khi bạn bắt đầu xây dựng skills một cách nghiêm túc, hai điều

 ---

-**Cập Nhật Lần Cuối**: Tháng 4 năm 2026
-**Phiên Bản Claude Code**: 2.1+
+**Cập Nhật Lần Cuối**: Ngày 25 tháng 8 năm 2026
+**Phiên Bản Claude Code**: 2.1.245
+**Nguồn**:
+- https://code.claude.com/docs/en/skills
 **Các Mô Hình Tương Thích**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -122,7 +122,7 @@ của tác nhân con để giải quyết vấn đề.
 | `tools` | Không | Danh sách phân tách bằng dấu phẩy của các công cụ cụ thể. Bỏ qua để kế thừa tất cả công cụ. Hỗ trợ cú pháp `Agent(agent_name)` để hạn chế tác nhân con có thể tạo |
 | `disallowedTools` | Không | Danh sách phân tách bằng dấu phẩy của các công cụ mà tác nhân con không được sử dụng |
 | `model` | Không | Mô hình để sử dụng: `sonnet`, `opus`, `haiku`, ID mô hình đầy đủ, hoặc `inherit`. Mặc định là mô hình tác nhân con được cấu hình |
-| `permissionMode` | Không | `default`, `acceptEdits`, `dontAsk`, `bypassPermissions`, `plan` |
+| `permissionMode` | Không | `manual` (đổi tên từ `default` ở v2.1.200 — tên cũ `default` vẫn được chấp nhận), `acceptEdits`, `dontAsk`, `bypassPermissions`, `plan`, `auto` |
 | `maxTurns` | Không | Số lượng lượt tác nhân tối đa mà tác nhân con có thể thực hiện |
 | `skills` | Không | Danh sách phân tách bằng dấu phẩy của các skills để tải trước. Tiêm toàn bộ nội dung skill vào bối cảnh của tác nhân con khi khởi động |
 | `mcpServers` | Không | MCP servers để cung cấp cho tác nhân con |
@@ -209,10 +209,10 @@ Claude Code bao gồm một số tác nhân con được tích hợp sẵn luôn
 |-------|-------|---------|
 | **general-purpose** | Kế thừa | Tác vụ đa bước phức tạp |
 | **Plan** | Kế thừa | Nghiên cứu cho chế độ lập kế hoạch |
-| **Explore** | Haiku | Khám phá codebase chỉ đọc (nhanh/trung bình/rất kỹ) |
-| **Bash** | Kế thừa | Lệnh terminal trong bối cảnh riêng |
-| **statusline-setup** | Sonnet | Cấu hình dòng trạng thái |
-| **Claude Code Guide** | Haiku | Trả lời câu hỏi về tính năng Claude Code |
+| **Explore** | Kế thừa (giới hạn ở Opus) | Khám phá codebase chỉ đọc (nhanh/trung bình/rất kỹ) |
+| **claude** | Kế thừa | Tác nhân dự phòng cho các tác vụ không khớp tác nhân chuyên biệt nào; có mọi công cụ khả dụng cho subagent. Cũng là tác nhân mặc định cho phiên nền được điều phối |
+| **statusline-setup** | Sonnet | Chạy khi bạn dùng `/statusline` để cấu hình dòng trạng thái |
+| **claude-code-guide** | Haiku | Trả lời câu hỏi về tính năng Claude Code |

 ### Tác Nhân General-Purpose

@@ -250,15 +250,15 @@ Claude Code bao gồm một số tác nhân con được tích hợp sẵn luôn
 - **"medium"** - Khám phá vừa phải, cân bằng tốc độ và sự kỹ lưỡng, cách tiếp cận mặc định
 - **"very thorough"** - Phân tích toàn diện qua nhiều vị trí và quy ước đặt tên, có thể mất nhiều thời gian hơn

-### Tác Nhân Bash
+### Tác Nhân Claude

 | Thuộc Tính | Giá Trị |
 |----------|-------|
 | **Mô Hình** | Kế thừa từ cha |
-| **Công Cụ** | Bash |
-| **Mục Đích** | Thực thi các lệnh terminal trong một cửa sổ bối cảnh riêng |
+| **Công Cụ** | Mọi công cụ khả dụng cho subagent |
+| **Mục Đích** | Tác nhân dự phòng cho các tác vụ không khớp tác nhân chuyên biệt nào |

-**Khi sử dụng**: Khi chạy các lệnh shell hưởng lợi từ bối cảnh cô lập.
+**Khi sử dụng**: Khi tác vụ không khớp với một tác nhân dựng sẵn chuyên biệt hơn. Đây cũng là tác nhân mặc định cho một phiên nền được điều phối; chế độ quyền mà nó khởi động phụ thuộc vào cách phiên đó được bắt đầu.

 ### Tác Nhân Cấu Trình Dòng Trạng Thái

@@ -1138,6 +1138,8 @@ graph TD

 ---

-**Cập Nhật Lần Cuối**: Tháng 4 năm 2026
-**Phiên Bản Claude Code**: 2.1+
+**Cập Nhật Lần Cuối**: Ngày 25 tháng 8 năm 2026
+**Phiên Bản Claude Code**: 2.1.245
+**Nguồn**:
+- https://code.claude.com/docs/en/sub-agents
 **Các Mô Hình Tương Thích**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -105,14 +105,28 @@ Giao thức Server-Sent Events đã lỗi thời để ủng hộ `http` nhưng
 claude mcp add --transport sse legacy-server https://example.com/sse
 ```

-### Giao Thức WebSocket / WebSocket Transport
+### Giao Thức WebSocket (`ws`) / WebSocket Transport (`ws`)

-Giao thức WebSocket cho các kết nối hai chiều liên tục:
+Máy chủ WebSocket giữ một kết nối hai chiều liên tục, phù hợp với các máy chủ MCP từ xa cần đẩy sự kiện tới Claude mà không cần được hỏi. Hãy dùng HTTP thay thế nếu máy chủ của bạn chỉ phản hồi các yêu cầu, vì HTTP hỗ trợ OAuth và cờ `claude mcp add --transport`, còn WebSocket thì không hỗ trợ cả hai.

-```bash
-claude mcp add --transport ws realtime-server wss://example.com/mcp
+Vì `--transport` không chấp nhận `ws`, hãy cấu hình trong `.mcp.json` hoặc qua `claude mcp add-json`:
+
+```json
+{
+  "type": "ws",
+  "url": "wss://mcp.example.com/socket",
+  "headers": {
+    "Authorization": "Bearer YOUR_TOKEN"
+  }
+}
 ```

+Mục `type: "ws"` chấp nhận cùng các trường `url`, `headers`, `headersHelper`, `timeout` và `alwaysLoad` như `http`. Xác thực **chỉ qua header** — không có luồng OAuth cho máy chủ WebSocket.
+
+> **Lưu ý**: Máy chủ WebSocket không xuất hiện trong đầu ra của `claude mcp list`. Hãy dùng `claude mcp get <tên>` hoặc bảng `/mcp` để kiểm tra chúng.
+
+Giống HTTP và SSE, kết nối WebSocket dùng cửa sổ nhàn rỗi 5 phút; stdio và WebSocket không có bộ đếm thời gian cho từng yêu cầu. Một mục `url` không có `type` sẽ báo lỗi và liệt kê `"http"`, `"sse"`, `"ws"` là các giá trị hợp lệ.
+
 ### Lưu Ý Cụ Thể Cho Windows / Windows-Specific Note

 Trên Windows native (không phải WSL), sử dụng `cmd /c` cho các lệnh npx:
@@ -1124,6 +1138,8 @@ export GITHUB_TOKEN="your_token"

 ---

-**Cập Nhật Lần Cuối**: Tháng 4 năm 2026
-**Phiên Bản Claude Code**: 2.1+
+**Cập Nhật Lần Cuối**: Ngày 25 tháng 8 năm 2026
+**Phiên Bản Claude Code**: 2.1.245
+**Nguồn**:
+- https://code.claude.com/docs/en/mcp
 **Các Mô Hình Tương Thích**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -57,8 +57,14 @@ Hooks được cấu hình trong các file settings với cấu trúc cụ thể
 | `hooks` | Mảng định nghĩa hook | `[{ "type": "command", ... }]` |
 | `type` | Loại hook: `"command"` (bash), `"prompt"` (LLM), `"http"` (webhook), `"mcp_tool"` (gọi công cụ MCP, từ v2.1.118), hoặc `"agent"` (subagent) | `"command"` |
 | `command` | Lệnh shell để thực thi | `"$CLAUDE_PROJECT_DIR/.claude/hooks/format.sh"` |
-| `timeout` | Timeout tùy chọn tính bằng giây (mặc định 60) | `30` |
+| `timeout` | Timeout tùy chọn tính bằng giây. Mặc định: 600 cho command/http/mcp_tool, 30 cho prompt, 60 cho agent. | `30` |
 | `once` | Nếu `true`, chạy hook chỉ một lần mỗi phiên | `true` |
+| `async` | Nếu `true`, chạy nền mà không chặn | `true` |
+| `asyncRewake` | Nếu `true`, chạy nền và đánh thức Claude khi mã thoát là 2. Ngầm bật `async`. | `true` |
+| `shell` | Chấp nhận `"bash"` hoặc `"powershell"`. Mặc định `"bash"`, hoặc `"powershell"` trên Windows khi chưa cài Git Bash. | `"bash"` |
+| `statusMessage` | Thông điệp spinner tùy chỉnh hiển thị khi hook đang chạy | `"Đang định dạng…"` |
+
+> **Lưu ý**: Một số sự kiện hạ thấp timeout mặc định. `UserPromptSubmit` hạ mặc định của command, http và mcp_tool xuống 30 giây, còn `MessageDisplay` hạ xuống 10 giây. Các hook `SessionEnd` dùng chung ngân sách 1,5 giây; nếu cài đặt của bạn khai báo `timeout` dài hơn cho một hook, Claude Code nâng ngân sách đó cho khớp, tối đa 60 giây.

 ### Các Mẫu Matcher / Matcher Patterns

@@ -157,6 +163,8 @@ Hooks xác thực dựa trên subagent mà spawn một agent chuyên dụng đ
 }
 ```

+> **Lưu ý**: Agent hooks là tính năng thử nghiệm và có thể thay đổi.
+
 **Các thuộc tính chính:**
 - `"type": "agent"` -- xác định đây là một agent hook
 - `"prompt"` -- mô tả nhiệm vụ cho subagent
@@ -517,6 +525,8 @@ Tất cả hooks nhận đầu vào JSON qua stdin:
 }
 ```

+> **`retry` (PermissionDenied)**: Dùng JSON `hookSpecificOutput.retry: true` để báo cho model biết nó có thể thử lại lệnh gọi công cụ đã bị từ chối.
+
 ## Các Biến Môi Trường / Environment Variables

 | Biến | Khả Dụng | Mô Tả |
@@ -866,8 +876,8 @@ Chỉnh sửa `~/.claude/settings.json` hoặc `.claude/settings.json` với c

 ---

-**Cập Nhật Lần Cuối**: Ngày 15 tháng 8 năm 2026
-**Phiên Bản Claude Code**: 2.1.233
+**Cập Nhật Lần Cuối**: Ngày 25 tháng 8 năm 2026
+**Phiên Bản Claude Code**: 2.1.245
 **Nguồn**:
 - https://code.claude.com/docs/en/hooks
 **Các Mô Hình Tương Thích**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -571,18 +571,15 @@ Claude: [Shows linter output from bg-5002]

 ### Configuration

-```json
-{
-  "backgroundTasks": {
-    "enabled": true,
-    "maxConcurrentTasks": 5,
-    "notifyOnCompletion": true,
-    "autoCleanup": true,
-    "logOutput": true
-  }
-}
+There is no `settings.json` block for background tasks. The feature is controlled by an environment variable:
+
+```bash
+# Turn background tasks off entirely
+export CLAUDE_CODE_DISABLE_BACKGROUND_TASKS=true
 ```

+Concurrency is not a background-task setting either — how many agents run at once is governed by `CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS` (default `20`).
+
 ---

 ## Scheduled Tasks
@@ -948,7 +945,7 @@ Claude Code supports keyboard shortcuts for efficiency. Here's the complete refe
 | `Ctrl+C` | Cancel current input/generation |
 | `Ctrl+D` | Exit Claude Code |
 | `Ctrl+G` | Edit plan in external editor |
-| `Ctrl+L` | Clear terminal screen |
+| `Ctrl+L` | Vẽ lại màn hình (chỉ repaint — phím tắt nhấn đúp để `/clear` đã bị gỡ ở v2.1.238) |
 | `Ctrl+O` | Toggle verbose output (view reasoning) |
 | `Ctrl+R` | Reverse search history |
 | `Ctrl+T` | Toggle task list view |
@@ -1428,7 +1425,7 @@ Connect external services for richer context:

 ### Permission modes in Desktop

-The Desktop App supports the same 4 permission modes as the CLI:
+The Desktop App supports the same permission modes as the CLI:

 | Mode | Behavior |
 |------|----------|
@@ -1873,6 +1870,6 @@ For more information about Claude Code and related features:

 ---

-**Cập Nhật Lần Cuối**: Tháng 4 năm 2026
-**Phiên Bản Claude Code**: 2.1+
+**Cập Nhật Lần Cuối**: Ngày 25 tháng 8 năm 2026
+**Phiên Bản Claude Code**: 2.1.245
 **Các Mô Hình Tương Thích**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
@@ -141,10 +141,10 @@ Các trợ lý AI chuyên biệt với context bị cô lập cho các tasks c
 |-------|-------------|-------|-------|-------------|
 | **general-purpose** | Tasks đa bước, nghiên cứu | All tools | Kế thừa model | Nghiên cứu phức tạp, tasks đa file |
 | **Plan** | Lập kế hoạch triển khai | Read, Glob, Grep, Bash | Kế thừa model | Thiết kế kiến trúc, lập kế hoạch |
-| **Explore** | Khám phá codebase | Read, Glob, Grep | Haiku 4.5 | Tìm kiếm nhanh, hiểu code |
-| **Bash** | Thực thi command | Bash | Kế thừa model | Operations Git, tasks terminal |
+| **Explore** | Khám phá codebase | Read, Glob, Grep | Kế thừa (giới hạn ở Opus) | Tìm kiếm nhanh, hiểu code |
+| **claude** | Tác nhân dự phòng cho các tác vụ không khớp tác nhân chuyên biệt nào | All tools | Kế thừa model | Tác vụ không có tác nhân chuyên biệt; tác nhân mặc định cho phiên nền được điều phối |
 | **statusline-setup** | Cấu hình status line | Bash, Read, Write | Sonnet 4.6 | Cấu hình hiển thị status line |
-| **Claude Code Guide** | Trợ giúp và tài liệu | Read, Glob, Grep | Haiku 4.5 | Nhận trợ giúp, học tính năng |
+| **claude-code-guide** | Trợ giúp và tài liệu | Read, Glob, Grep | Haiku 4.5 | Nhận trợ giúp, học tính năng |

 ### Subagent Configuration Fields

@@ -11,7 +11,6 @@ Chúng tôi cung cấp các bản cập nhật bảo mật cho các phiên bản
 | Phiên Bản | Trạng Thái | Hỗ Trợ Đến |
 |---------|--------|---------------|
 | Mới nhất (main) | ✅ Hoạt Động | Hiện tại + 6 tháng |
-| Bản phát hành 1.x | ✅ Hoạt Động | Đến phiên bản lớn tiếp theo |

 **Lưu ý**: Là một dự án hướng dẫn giáo dục, chúng tôi tập trung vào việc duy trì các thực hành tốt nhất hiện tại và bảo mật tài liệu hơn là hỗ trợ phiên bản truyền thống. Các bản cập nhật được áp dụng trực tiếp vào nhánh chính.

@@ -87,4 +86,7 @@ Thank you for helping keep Claude How To secure!

 ---

-**Cập Nhật Lần Cuối**: Tháng 4 năm 2026
+**Cập Nhật Lần Cuối**: Ngày 25 tháng 8 năm 2026
+**Phiên Bản Claude Code**: 2.1.245
+**Nguồn**:
+- https://code.claude.com/docs/en/overview
@@ -65,9 +65,9 @@ Claude Code 自带一些常见角色，例如：
 - 通用助手
 - Planning agent
 - Explore agent
- Bash agent
+- claude 兜底 agent
 - Statusline 相关 agent
- Claude Code Guide agent
+- claude-code-guide agent

 ## 管理 Subagents

@@ -140,10 +140,10 @@ Claude Code 提供 6 种权限模式，用来控制工具调用如何被授权
 |-------|-------------|-------|-------|-------------|
 | **general-purpose** | 多步任务、研究 | 所有工具 | 继承当前模型 | 复杂研究、多文件任务 |
 | **Plan** | 实现规划 | Read、Glob、Grep、Bash | 继承当前模型 | 架构设计、规划 |
-| **Explore** | 代码库探索 | Read、Glob、Grep | Haiku 4.5 | 快速搜索、理解代码 |
-| **Bash** | 命令执行 | Bash | 继承当前模型 | git 操作、终端任务 |
+| **Explore** | 代码库探索 | Read、Glob、Grep | 继承（上限为 Opus） | 快速搜索、理解代码 |
+| **claude** | 不适合更专门 agent 的任务的通用兜底 agent | 所有工具 | 继承当前模型 | 没有专门 agent 的任务；被调度的后台会话的默认 agent |
 | **statusline-setup** | 状态栏配置 | Bash、Read、Write | Sonnet 4.6 | 配置状态栏显示 |
-| **Claude Code Guide** | 帮助与文档 | Read、Glob、Grep | Haiku 4.5 | 获取帮助、学习功能 |
+| **claude-code-guide** | 帮助与文档 | Read、Glob、Grep | Haiku 4.5 | 获取帮助、学习功能 |

 ### Subagent 配置字段

@@ -878,8 +878,8 @@ MIT 许可证，详见 [LICENSE](LICENSE)。你可以自由使用、修改和分

 ---

-**最后更新**：2026 年 8 月 15 日
-**Claude Code 版本**：2.1.233
+**最后更新**：2026 年 8 月 25 日
+**Claude Code 版本**：2.1.245
 **来源**：
 - https://code.claude.com/docs/en/hooks
 **兼容模型**：Claude Sonnet 4.6、Claude Opus 4.6、Claude Haiku 4.5
@@ -11,7 +11,6 @@ Claude How To 项目的安全对我们非常重要。本文档说明我们的安
 | 版本 | 状态 | 支持至 |
 |---------|--------|---------------|
 | 最新版（main） | ✅ 活跃 | 当前版本 + 6 个月 |
-| 1.x 发布版 | ✅ 活跃 | 直到下一个大版本 |

 **注意**：作为一个教育指南项目，我们更关注维护当前最佳实践和文档安全，而不是传统意义上的版本支持。更新会直接应用到 main 分支。

@@ -326,7 +325,10 @@ pip-audit
 - 项目范围发生变化时
 - 至少每年一次

-**最后更新**：2026 年 1 月
+**最后更新**：2026 年 8 月 25 日
+**Claude Code 版本**：2.1.245
+**来源**：
+- https://code.claude.com/docs/en/overview
 **下次审查**：2027 年 1 月

 ---
@@ -597,12 +597,13 @@ docs(skills): Add comprehensive code review skill

 ```markdown
 ---
-**Last Updated**: March 2026
-**Claude Code Version**: 2.1.97
+**Last Updated**: August 2026
+**Claude Code Version**: 2.1.245
 **Compatible Models**: Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5
 ```

- 使用“月份 + 年份”格式，例如 “March 2026”
+- 使用“月份 + 年份”格式，例如 “August 2026”
+- 使用当前同步运行捕获的版本号，而非此处示例中显示的值
 - 功能变化时更新版本
 - 列出所有兼容模型

@@ -631,7 +632,7 @@ docs(skills): Add comprehensive code review skill

 ---

-**最后更新**: 2026 年 8 月 15 日
-**Claude Code 版本**: 2.1.233
+**最后更新**: 2026 年 8 月 25 日
+**Claude Code 版本**: 2.1.245
 **来源**:
 - https://code.claude.com/docs/en/overview
@@ -1512,16 +1512,14 @@ claude -r "Feature"
    "autoEnter": true,
    "requireApproval": true
  },
-  "backgroundTasks": {
-    "enabled": true,
-    "maxConcurrentTasks": 5
-  },
  "permissions": {
    "defaultMode": "manual"
  }
 }
 ```

+`settings.json` 中没有后台任务配置块 —— 该功能由环境变量 `CLAUDE_CODE_DISABLE_BACKGROUND_TASKS` 控制，并发数由 `CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS`（默认 `20`）决定。
+
 **详见：** [09-advanced-features/README.md](09-advanced-features/README.md)

 ---