メニュー
Pull request review guidance for correctness, security, tests, performance, API contract drift, migrations, and error handling. Use when reviewing diffs, pull requests, or risky refactors.
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。
Use when building or debugging apps that call the Claude API — implementing tool use, streaming, vision, prompt caching, batch processing, extended thinking, or an agentic loop with the Anthropic SDK.
Use when configuring Claude Code — installing skills or agents, writing hook configurations, setting up tool permissions, registering MCP servers, or authoring CLAUDE.md project instructions.
Use when writing Dockerfiles, setting up docker-compose for local dev, configuring Kubernetes resources (Deployment, Service, Ingress, HPA), sizing pod resource limits, or packaging a service with Helm.
Use when designing a schema, adding indexes to fix slow queries, writing a zero-downtime migration, diagnosing N+1 issues with EXPLAIN ANALYZE, or configuring connection pooling for a PostgreSQL-backed service.
End-to-end (E2E) testing patterns for user journeys, browser automation, UI state, and production-like environment validation. Use when testing the system as a black box from the user's perspective.
Use when adding feature flag support to a service, designing a percentage-based rollout, setting up A/B experiments or multivariate tests, choosing between LaunchDarkly, Unleash, and OpenFeature, writing tests for flag-gated code, or managing flag lifecycle and cleanup.
SOC 職業分類に基づく
| name | code-review |
| description | Pull request review guidance for correctness, security, tests, performance, API contract drift, migrations, and error handling. Use when reviewing diffs, pull requests, or risky refactors. |
Review code for correctness first, then for safety, operability, and maintainability.
| Priority | Check |
|---|---|
| 1 | Logic correctness and regression risk |
| 2 | Security, privacy, and data integrity |
| 3 | Test coverage and validation gaps |
| 4 | Performance, scaling, and resource usage |
| 5 | Style, naming, and cleanup |
| Comment Type | Use When | Prefix |
|---|---|---|
| Blocking | Incorrect behavior, security hole, migration risk, broken contract | Blocking: |
| Suggestion | Readability improvement, minor cleanup, non-critical refactor | Suggestion: |
| Question | Intent or assumption is unclear | Question: |
| Area | Questions |
|---|---|
| Logic | Are edge cases, retries, and failure paths handled? |
| Security | Are auth checks, secrets, and untrusted inputs handled safely? |
| Tests | Do tests cover new behavior and likely regressions? |
| Performance | Does this add N+1 queries, repeated allocations, or hot-loop work? |
| API contracts | Will clients or jobs break on shape changes? |
| Migrations | Is rollout safe across old and new code versions? |
| Error handling | Are failures explicit, typed, and observable? |
| Preferred | Avoid |
|---|---|
| Direct, specific, file-scoped feedback | Vague comments like "this feels wrong" |
| Explain impact | Nitpicks framed as blockers |
| Offer a fix path when practical | Personal or emotional language |
BAD
This code is bad. Please redo.
GOOD
Blocking: [payments/service.ts] charge retries on 500 but not on timeout, so transient network failures can still drop payments. Add retry handling for socket timeout errors and cover it with a test.