Content
56%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body delivers a clear, well-sequenced TDD workflow with mostly executable, project-specific examples (test patterns, mocks, coverage config). Its main weaknesses are verbosity — generic TDD principles and best-practice lists Claude already knows — and zero progressive disclosure, with all patterns inlined in a single long file instead of split into reference files.
Suggestions
Cut or compress sections that restate common knowledge — the red-green-refactor explanation, the generic 'ベストプラクティス' list, and boilerplate Jest/Playwright scaffolding — keeping only the project-specific patterns (Supabase/Redis/OpenAI mocks, coverage thresholds, file layout).
Split the detailed test patterns (E2E Playwright examples, per-service mock recipes, common mistakes) into references/ files (e.g. references/test-patterns.md, references/mocks.md) and keep SKILL.md as a concise workflow overview with well-signaled links.
Complete the stub examples in the workflow steps (Steps 2 and 4) or replace them with a pointer to the full patterns, and add an explicit error-recovery step (e.g. 'if tests still fail or coverage < 80%, revisit implementation before proceeding').
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Substantial sections restate knowledge Claude already has: "常にテストを最初に書き、次にテストに合格するコードを実装します" (basic red-green-refactor), the generic "ベストプラクティス" list (Arrange-Act-Assert, test edge cases, keep tests under 50ms), and full boilerplate Jest/Playwright patterns. This matches 'noticeably verbose; several unnecessary explanations or padded sections' rather than the level-3 'some unnecessary explanation'. | 2 / 5 |
Actionability | Most guidance is copy-paste executable: the complete Button unit test, the GET /api/markets integration test, the Supabase/Redis/OpenAI mock recipes, npm test / npm run test:coverage commands, and the jest coverageThresholds config. It falls short of 5 because core workflow Steps 2 and 4 are empty stubs ("// テスト実装", "// 実装はここ") and the database-error test body is unwritten. | 4 / 5 |
Workflow Clarity | The 7-step TDD sequence is clearly ordered with explicit checkpoints: Step 3 "テストを実行（失敗するはず）", Step 5 re-run expecting green, and Step 7 "カバレッジを確認" against an 80% threshold. Not a 5 because there is no explicit error-recovery feedback loop (no instruction for what to do when tests still fail or coverage falls short). | 4 / 5 |
Progressive Disclosure | No bundle files exist, and all ~410 lines are inlined in SKILL.md with no references. Section headers give reasonable structure, but content that clearly belongs in separate files — full E2E patterns, per-service mock recipes, and the common-mistakes gallery — is inline, matching 'content that should be separate is inline'. Not 2 because organization and headers are good, not minimal. | 3 / 5 |
Total | 13 / 20 Passed |