Florian BRUNIAUX
|
4a0a0bf30e
|
docs: complete factual audit pass 2 — 90+ corrections
Second 10-agent parallel audit covering all remaining sections:
ultimate-guide.md (ch1-ch11), workflows/ (17 files), quiz/ (12 files),
examples/agents+skills+commands. Source of truth: official Anthropic docs.
Key corrections:
Hook system (+8 missing events):
- Complete 17-event list: PermissionRequest, PostToolUseFailure, SubagentStart,
TeammateIdle, TaskCompleted, WorktreeCreate, WorktreeRemove, SessionEnd
- SessionStart confirmed valid (previous audit wrongly doubted it)
- Hook output format: hookSpecificOutput.permissionDecision (not {"decision":"block"})
- Missing common input fields added: transcript_path, cwd, permission_mode
Agent YAML frontmatter (13 valid fields restored/added):
- Restored: disallowedTools, memory, background, isolation, skills, permissionMode, hooks
- Added new: maxTurns, mcpServers
- Fixed: tools format is comma-separated (not space-separated)
Plan Mode (12 occurrences fixed):
- Ctrl+G = "open plan in text editor" (NOT "enter plan mode")
- Plan Mode = Shift+Tab × 2 (Normal → acceptEdits → plan)
Commands table (10.1) + built-in commands (6.1):
- Added 18+ missing commands: /copy, /doctor, /hooks, /memory, /model,
/config, /permissions, /remote-control, /rename, /resume, /sandbox, etc.
Workflow files:
- agent-teams.md: removed fake --experimental-agent-teams flag
- hooks.yaml + post_edit event → settings.json + PostToolUse (2 files)
- TodoWrite → TaskCreate/TaskUpdate (3 files)
- task-management.md: removed fake "failed" task status
Quiz / examples:
- 01-010: Esc stops mid-action (not Ctrl+C)
- refactoring-specialist.md: removed MultiEdit (not a valid tool)
- ast-grep-patterns.md: name field (not title)
- validate-changes.md, diagnose.md: field name fixes
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
2026-02-26 18:21:28 +01:00 |
|
Florian BRUNIAUX
|
8a4d116e2e
|
feat(docs): add LLM Handbook + Google Whitepaper integration v3.3.0
Advanced Guardrails:
- prompt-injection-detector.sh (PreToolUse)
- output-validator.sh (PostToolUse heuristics)
- claudemd-scanner.sh (SessionStart injection detection)
- output-secrets-scanner.sh (PostToolUse secrets leak prevention)
Observability & Monitoring:
- session-logger.sh (JSONL activity logging)
- session-stats.sh (cost tracking & analysis)
- guide/observability.md (full documentation)
LLM-as-a-Judge Evaluation:
- output-evaluator.md agent (Haiku)
- /validate-changes command
- pre-commit-evaluator.sh (opt-in git hook)
Google Agent Whitepaper Integration:
- Context Triage Guide (Section 2.2.4)
- CLAUDE.md Injection Warning (Section 3.1.3)
- Agent Validation Checklist (Section 4.2.4)
- MCP Security: Tool Shadowing & Confused Deputy (Section 8.6)
- Session vs Memory patterns (Section 3.3.3)
Stats: 10 new files, 8 modified, 5 new guide sections
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
|
2026-01-14 21:00:49 +01:00 |
|