Use when creating or revising ARCHITECTURE.md — a stable project knowledge document for newcomer onboarding, contributor reference, and AI agent context. Skip for projects under ~1k lines where CLAUDE.md alone suffices, or for user-facing docs (README, guides).
Perform comprehensive project health checks with fast (read-only) and full (build/test/dependency) modes, generating detailed Markdown reports with actionable suggestions
Use when performing a work process retrospective — analyzes plan vs actual, detects cross-session patterns in .claude/retro-history.md, generates structured reports, and creates GitHub issues for continuous improvement. Triggers on "retro", "retrospective", "回顾", "复盘". Manual trigger only via /retro.
Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup
Use when executing small, logically straightforward tasks that don't require design docs or detailed plans. Quick analyze-confirm-execute cycle for simple fixes, small features, or setup tasks.
Internal fallback for plan execution when subagents are unavailable — invoked automatically by writing-plans, not intended for direct user selection
Use when executing implementation plans with independent tasks in the current session
Use when deploying an application and verifying the deployment works — reads DEPLOY.md, executes deployment, runs smoke tests or E2E tests, reports structured results. Triggers on "deploy", "deploy and verify", "上线", "验证", or after finishing a development branch.