ワンクリックで
code-review-excellence
Meta-level code review guidance — how to review well, not just what to check
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。
メニュー
Meta-level code review guidance — how to review well, not just what to check
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
Structured brainstorming for exploring new features or solving complex problems
Scope, architect, and plan new tasks
Multi-angle code review before merging
Summarize the day's work and put it in context
Wrap up work — atomic commits and push to origin
Full end-to-end (E2E) testing skill utilizing the browser. Make sure to use this skill whenever the user asks to "test the app", "run end to end tests", "verify the UI", "check if the app works", or wants you to use the browser to interact with and test the application flow. This skill orchestrates creating a test plan, getting user approval, navigating the application using the browser, independently fixing obstacles, and generating a progressive test report.
| name | code-review-excellence |
| description | Meta-level code review guidance — how to review well, not just what to check |
How to conduct a high-quality code review. This skill provides meta-level guidance for the /review workflow — the domain-specific checklists (security, performance, architecture, accessibility) tell you what to check; this skill tells you how to review well.
| Substance (flag) | Style (skip) |
|---|---|
| Logic errors | Formatting (Pint/ESLint handles this) |
| Missing edge cases | Variable naming preferences |
| Security gaps | Blank line placement |
| Architectural violations | Import ordering |
| Missing tests | Comment phrasing |
If a linter or formatter can catch it, don't spend review time on it.
| ❌ Vague | ✅ Actionable |
|---|---|
| "This could be better" | "Extract lines 42-58 into a calculateDiscount() method — it's reused in OrderAction too" |
| "Performance concern here" | "This User::all() loads ~5K records into memory. Use paginate(50) or a scoped query" |
| "Needs tests" | "Add a test for the case where discount > subtotal — currently returns a negative total" |
🔴 Must: — blocks merging. Security, correctness, data integrity.🟡 Should: — strong recommendation. Missing tests, standards violations.🟢 Nit: — optional improvement. Take it or leave it.❓ Question: — genuine question, not a disguised demand your way.cursorPaginate() here — much better than offset pagination for this dataset."For the /review workflow, follow this sequence: