chore: sync docs and apply audit improvements (#17)

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

Author: github-actions[bot]

agents/code-reviewer.md

@@ -1,6 +1,6 @@
 ---
 name: code-reviewer
-description: Senior engineer code reviewer for quality, maintainability and readability
+description: Senior engineer code reviewer for quality, maintainability and readability. Use proactively after code changes, when asked to "review code", "review changes", or "check the implementation". Do NOT use for reviewing PRs from GitHub (use review-pr skill instead).
 tools: Read, Grep, Glob, Bash
 skills:
   - architect

agents/code-simplifier.md

@@ -1,6 +1,6 @@
 ---
 name: code-simplifier
-description: Code refinement specialist for clarity, consistency and maintainability
+description: Code refinement specialist for clarity, consistency and maintainability. Use proactively after implementation phases, when asked to "simplify", "clean up", "polish", or "refactor" code. Do NOT change behavior or add features — only improve how code reads and is organized.
 disallowedTools: Agent
 maxTurns: 50
 ---

agents/symfony-simplifier.md

@@ -1,6 +1,6 @@
 ---
 name: symfony-simplifier
-description: Code refinement specialist for Symfony applications
+description: Code refinement specialist for Symfony/PHP applications. Use proactively after Symfony implementation phases, when asked to "simplify", "clean up", "polish", or "refactor" PHP/Symfony code. Preferred over code-simplifier for PHP/Symfony projects. Do NOT change behavior or add features.
 disallowedTools: Agent
 maxTurns: 50
 ---

commands/resolve.md

@@ -1,7 +1,7 @@
 ---
 description: Main orchestrator for ticket resolution workflow - fetch, analyze, plan, implement, simplify, review, PR
 argument-hint: [ticket-id] [--auto] [--continue] [--refine-plan] [--plan-only] [--init] [--source youtrack|github] [--skip-simplify] [--skip-review] [--pr] [--draft]
-allowed-tools: Read, Glob, Grep, Bash, Write, Edit, Task, AskUserQuestion, mcp__youtrack__get_issue, mcp__youtrack__get_issue_comments, mcp__youtrack__get_issue_attachments, mcp__figma-screenshot__figma_screenshot
+allowed-tools: Read, Glob, Grep, Bash, Write, Edit, Agent, AskUserQuestion, mcp__youtrack__get_issue, mcp__youtrack__get_issue_comments, mcp__youtrack__get_issue_attachments, mcp__figma-screenshot__figma_screenshot
 ---
 
 # /resolve - Ticket Resolution Workflow

commands/review-code.md

@@ -1,7 +1,7 @@
 ---
 description: Perform code review on current branch changes via isolated fresh agent
 argument-hint: [--ticket <id>] [--base <branch>] [--fix] [--severity <level>]
-allowed-tools: Read, Glob, Grep, Bash, Edit, Write, AskUserQuestion, Task
+allowed-tools: Read, Glob, Grep, Bash, Edit, Write, AskUserQuestion, Agent
 ---
 
 # CODE REVIEW COMMAND

docs/changelog-last-version.txt

@@ -1 +1 @@
-v2.1.72
+v2.1.77

docs/claude-cookbook-best-practices.md

@@ -1,9 +1,9 @@
 # Claude Code Best Practices
 
 > Document de reference pour Claude Code CLI
-> Last updated: 2026-03-10 (sync #5)
+> Last updated: 2026-03-17 (sync #7)
 > Use `/sync-cookbook` to refresh this document
-> Latest version: v2.1.72
+> Latest version: v2.1.77
 
 ---
 
@@ -69,6 +69,8 @@ Context and instructions for Claude at multiple levels:
 - Check CLAUDE.md into git so the team can contribute
 - Treat like code: review when things go wrong, test changes by observing behavior
 
+**HTML comments hidden (v2.1.72):** `<!-- ... -->` in CLAUDE.md are hidden from Claude when auto-injected (not counted in context). Comments remain visible when the file is read with the Read tool.
+
 **Include / Exclude:**
 
 | Include | Exclude |
@@ -147,6 +149,10 @@ Context and instructions for Claude at multiple levels:
 | `awsAuthRefresh` | string | Custom script for AWS SSO login |
 | `awsCredentialExport` | string | Custom script that outputs JSON with AWS credentials |
 | `otelHeadersHelper` | string | Script for dynamic OpenTelemetry headers |
+| `modelOverrides` | object | Map model picker entries to custom provider IDs (Bedrock ARNs, etc.) |
+| `autoMemoryDirectory` | string | Custom directory for auto-memory storage (default: `~/.claude/memory`) |
+| `effortLevel` | string | Persist effort level across sessions: `"low"`, `"medium"`, `"high"`. Set via `/effort`. Applies to Opus 4.6 and Sonnet 4.6 |
+| `feedbackSurveyRate` | number | **Managed**: Session quality survey sample rate (0-1, enterprise only) |
 
 ### MCP Server Settings
 
@@ -190,6 +196,7 @@ Context and instructions for Claude at multiple levels:
     "allowUnsandboxedCommands": true,
     "filesystem": {
       "allowWrite": ["//tmp/build", "~/.kube"],
+      "allowRead": ["~/.kube"],
       "denyWrite": ["//etc"],
       "denyRead": ["~/.aws/credentials"]
     },
@@ -210,6 +217,21 @@ Context and instructions for Claude at multiple levels:
 
 **Path prefixes:** `//` = absolute, `~/` = home, `/` = relative to settings file, `./` = relative
 
+### Worktree Configuration
+
+```json
+{
+  "worktree": {
+    "sparsePaths": ["src/", "tests/", ".claude/"],
+    "symlinkDirectories": ["node_modules", ".venv"]
+  }
+}
+```
+
+**`sparsePaths`:** In large monorepos, check out only necessary directories via git sparse-checkout when using `claude --worktree`.
+
+**`symlinkDirectories`:** Directories to symlink from main repo into each worktree (reduces disk usage for large dependency directories).
+
 ### Environment Variables
 
 | Variable | Purpose |
@@ -219,7 +241,7 @@ Context and instructions for Claude at multiple levels:
 | `ANTHROPIC_DEFAULT_HAIKU_MODEL` | Haiku model override |
 | `ANTHROPIC_DEFAULT_OPUS_MODEL` | Opus model override |
 | `CLAUDE_CODE_EFFORT_LEVEL` | `low`, `medium`, `high` (thinking budget) |
-| `CLAUDE_CODE_MAX_OUTPUT_TOKENS` | Max output tokens (default: 32000, max: 64000) |
+| `CLAUDE_CODE_MAX_OUTPUT_TOKENS` | Max output tokens (default: 64000 for Opus 4.6, 32000 for Sonnet, upper bound: 128000 for Opus/Sonnet 4.6) (v2.1.77) |
 | `CLAUDE_CODE_AUTOCOMPACT_PCT_OVERRIDE` | Context % for auto-compaction (1-100) |
 | `CLAUDE_CODE_SIMPLE` | Minimal mode (Bash, file tools only) |
 | `CLAUDE_CODE_SHELL` | Override shell detection |
@@ -236,6 +258,7 @@ Context and instructions for Claude at multiple levels:
 | `CLAUDE_CODE_DISABLE_1M_CONTEXT` | Disable 1M context window |
 | `CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING` | Disable adaptive reasoning for Opus/Sonnet 4.6 |
 | `CLAUDE_CODE_DISABLE_CRON` | Disable scheduled tasks (cron tools unavailable) |
+| `CLAUDE_CODE_SESSIONEND_HOOKS_TIMEOUT_MS` | Timeout for `SessionEnd` hooks on exit (v2.1.74) |
 | `CLAUDE_CODE_FILE_READ_MAX_OUTPUT_TOKENS` | Override default token limit for file reads |
 | `CLAUDE_CODE_ENABLE_TASKS` | Set to `false` to revert to TODO list instead of task tracking |
 | `CLAUDE_CODE_ENABLE_PROMPT_SUGGESTION` | Set to `false` to disable prompt suggestions |
@@ -294,6 +317,9 @@ Installable extensions from marketplace bundling skills, hooks, subagents, and M
 | `WorktreeCreate` | Creation worktree | Custom git behavior | Pas de matcher |
 | `WorktreeRemove` | Suppression worktree | Cleanup | Pas de matcher |
 | `PreCompact` | Avant compaction | Pre-traitement | `manual`, `auto` |
+| `PostCompact` | Apres compaction | React to compacted state, log | `manual`, `auto` |
+| `Elicitation` | MCP elicitation dialog | Override MCP input | Nom serveur MCP |
+| `ElicitationResult` | Reponse a elicitation MCP | Valider/bloquer reponse | Nom serveur MCP |
 | `SessionEnd` | Fin session | Cleanup, rapport | `clear`, `logout`, `prompt_input_exit`, `bypass_permissions_disabled`, `other` |
 
 ### Hook Types
@@ -373,6 +399,24 @@ Note: En v2.1.63, `Task` a ete renomme en `Agent`. Les references `Task(...)` ex
 }
 ```
 
+**PostToolUse (MCP tools only)** - remplacer la sortie de l'outil:
+
+```json
+{
+  "hookSpecificOutput": {
+    "hookEventName": "PostToolUse",
+    "updatedMCPToolOutput": [{"type": "text", "text": "Remapped output"}]
+  }
+}
+```
+
+**WorktreeCreate** - retourner le chemin absolu pour VCS personnalise (remplace la creation git standard):
+
+```bash
+# stdout = chemin absolu du worktree cree
+echo "/path/to/custom/worktree"
+```
+
 **PermissionRequest** - via `hookSpecificOutput.decision`:
 
 ```json
@@ -704,9 +748,19 @@ Commands livrees avec Claude Code:
 | `/batch` | Batch operations on multiple files |
 | `/claude-api` | Build apps with Claude API or Anthropic SDK |
 | `/security-review` | Security review of current changes |
-| `/copy` | Interactive picker: select a code block or the full response to copy (v2.1.59) |
+| `/copy` | Copy responses: `/copy` for picker, `/copy N` for Nth-latest response (v2.1.77) |
+| `/effort` | Set model effort level (`low`, `medium`, `high`) for thinking budget (v2.1.76) |
 | `/loop` | Run a prompt or slash command on a recurring interval (e.g., `/loop 5m check deploy`) (v2.1.71) |
 
+### /plan with Optional Description (v2.1.72)
+
+```bash
+/plan                          # Enter plan mode
+/plan fix the auth bug         # Enter plan mode with immediate description
+```
+
+The optional argument sets an initial description so Claude starts planning immediately without waiting for context.
+
 ### Cron Scheduling Tools (New in v2.1.71)
 
 Schedule prompts or commands to run on recurring intervals **within a session**. Useful for monitoring tasks, periodic checks, or recurring reminders.
@@ -932,7 +986,7 @@ When enabled: memory directory instructions added to system prompt, first 200 li
 
 - **Foreground** (default): Blocks main conversation. Permission prompts passed through.
 - **Background**: Runs concurrently. Permissions pre-approved upfront. Auto-denies unapproved actions. `Ctrl+B` to background a running task.
-- **Resume**: Each subagent gets an `agent_id`. Ask Claude to continue previous work to resume with full context preserved.
+- **Resume**: Each subagent gets an `agent_id`. Use `SendMessage({to: agentId})` to continue a previously spawned agent (Agent tool no longer accepts a `resume` parameter since v2.1.77). `SendMessage` auto-resumes stopped agents in the background instead of returning an error.
 
 ### Subagent Transcripts & Resume
 
@@ -985,6 +1039,7 @@ Coordination automatisee de multiples sessions avec taches partagees, messaging,
 | `--disallowedTools` | Block specific tools |
 | `--agent <name>` | Start with a custom agent |
 | `--agents <json>` | Define session-only subagents as JSON |
+| `-n`, `--name <name>` | Set display name for session at startup (v2.1.76) |
 | `--worktree` | Run in isolated git worktree |
 
 ### Slash Commands
@@ -1015,7 +1070,7 @@ Coordination automatisee de multiples sessions avec taches partagees, messaging,
 | `/batch` | Batch operations |
 | `/loop` | Run prompt/command on recurring interval (e.g., `/loop 5m check deploy`) |
 | `/debug` | Toggle debug logging mid-session |
-| `/fork` | Fork conversation into separate session |
+| `/branch` | Fork conversation into separate session (`/fork` still works as alias) (v2.1.77) |
 | `/color` | Change session color (`/color default` to reset) |
 | `/terminal-setup` | Install terminal shortcuts (Shift+Enter for iTerm2/VS Code) |
 | `/security-review` | Security review of current changes |
@@ -1194,7 +1249,7 @@ cd ../feature-branch
 claude  # Session separee, contexte preserve
 ```
 
-Ou utiliser `claude --worktree` pour isolation automatique.
+Ou utiliser `claude --worktree` pour isolation automatique. L'outil `ExitWorktree` (v2.1.72) permet de quitter une session `EnterWorktree` depuis le REPL.
 
 ### Parallel Sessions
 

settings.json

@@ -40,6 +40,10 @@
       "WebFetch(domain:shipyard.build)",
       "WebFetch(domain:awesomeclaude.ai)"
     ],
+    "ask": [
+      "Bash(git push *)",
+      "Bash(cd * && git push *)"
+    ],
     "deny": [
       "Bash(rm -rf *)",
       "Bash(rm -rf /)",

skills/code-review/SKILL.md

@@ -46,7 +46,7 @@ Required context (gathered by YOU, passed inline to the agent):
 ### Step 2: Delegate to Fresh Agent
 
 ```yaml
-Task:
+Agent:
   subagent_type: code-reviewer
   description: "Fresh code review {ticket-id}"
   prompt: |