メニュー
Full 10-point expert code review. Use before any PR or MR. Checks correctness, cross-layer consistency, transaction safety, test coverage, and security. Spawns reviewer subagent for isolation.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Build the project and verify it compiles cleanly. Use after making changes to confirm nothing is broken before running tests. Reads the build command from context/build.md.
Auto-configure .github/copilot-instructions.md and context/ knowledge docs for a new project. Use when setting up Brain Bootstrap in a new repository. Discovers the codebase, fills in templates, and writes project-specific configuration.
Save session state before context gets full or before ending. Writes current task state, branch, and loaded docs to context/tasks/todo.md so the next session can resume cleanly.
Clean workspace — build artifacts, dependencies, caches, Docker volumes, or temp files. Accepts arguments like build, deps, all, cache, docker, tasks, reinstall.
Load all relevant context/ knowledge files for a domain area. Use at session start or when switching to a new topic. Accepts a domain keyword like api, database, auth, build, security.
Query the database — list schemas, tables, describe a table, or run SQL. Accepts arguments like schemas, tables, describe <table>, or raw SQL.
| name | review |
| description | Full 10-point expert code review. Use before any PR or MR. Checks correctness, cross-layer consistency, transaction safety, test coverage, and security. Spawns reviewer subagent for isolation. |
Run a full expert code review before merging.
git --no-pager diff main...HEAD | head -500
Or specify a branch: git --no-pager diff <base>...<head> | head -500
context/templates.md for PR template requirementscontext/rules.md for project-specific rules"Spawn a reviewer subagent to review this PR. Diff: [paste diff]. Focus on the 10-point protocol."
Point 1 — Ticket re-read: Does every scenario in the ticket/task get addressed?
Point 2 — Cross-layer consistency: Grep every new field name, type, enum across ALL layers:
grep -rn --color=never '<new_field_name>' . | head -50
Point 3 — Enum completeness: Are all enum values handled in every switch/if-chain?
Point 4 — Transaction safety: Are there external calls (HTTP, queue, email) inside DB transactions? → Move them outside.
Point 5 — Race conditions: Is there shared mutable state accessed concurrently without synchronization?
Point 6 — Test coverage: Every new branch/case/error path has a dedicated test?
Point 7 — Pre-existing vs. introduced: Run lint before and after. Distinguish new warnings from pre-existing ones.
Point 8 — Cross-branch merge safety: Will this still apply cleanly after other pending branches merge?
Point 9 — Security: New HTTP calls? Data exposed? New permissions? Unsafe defaults?
Point 10 — Confidence gate: Overall confidence 0-100. Only proceed to PR description if ≥95.
Report format:
Use /mr skill after the review passes.