Content
81%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.
A well-structured TDD skill: a high-value project-context prelude with exact commands and enforcement details, a clearly sequenced workflow with explicit RED/GREEN validation loops and checklists, and well-signaled one-level references. Weaknesses are minor: some conceptual padding in the Philosophy section, no inline test-writing example, and referenced sidecars missing from the bundle.
Suggestions
Tighten the Philosophy section — the good/bad test paragraphs and the four-bullet 'crap tests' list largely restate what the workflow rules already imply.
Inline one minimal passing/failing test example (e.g. a single vitest *.test.ts case) so the core action doesn't depend entirely on the tests.md sidecar.
Ship the five referenced sidecars (tests.md, mocking.md, refactoring.md, deep-modules.md, interface-design.md) with the skill bundle, or remove the links — as delivered, every reference dead-ends.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Largely efficient: the project-context prelude is dense and non-obvious, and the workflow/checklist sections carry no padding. Minor over-explanation in the Philosophy section ("Good tests are integration-style... Bad tests are coupled to implementation") and the four-bullet 'crap tests' rationale, which restate ideas Claude can infer. Not 5 because those passages could be tightened; not 3 because nothing is generic filler. | 4 / 5 |
Actionability | Concrete, executable specifics throughout: "pnpm turbo test --filter=<workspace>", "*.test.ts" vs "*.spec.ts" conventions, "PUSH_WITHOUT_TESTS=1" and the "tdd-exempt" label escape hatches, plus exact file paths (scripts/guards.config.sh, docs/domain-glossary.md). Not 5 because the core skill action — how to write a good test — is entirely deferred to referenced sidecars with no inline example. Not 3 because what is inline is specific and executable, not pseudocode. | 4 / 5 |
Workflow Clarity | Clear sequence (Planning → Tracer Bullet → Incremental Loop → Refactor) with explicit validation checkpoints: each RED→GREEN cycle is a validate-then-fix loop, "Run tests after each refactor step", "Never refactor while RED. Get to GREEN first", plus a per-cycle checklist and planning checklists with user confirmation gates. Matches the anchor-5 pattern (explicit validation, feedback loops, checklists); no destructive/batch cap applies. | 5 / 5 |
Progressive Disclosure | Good structure: overview-sized body with clearly signaled, one-level-deep references ([tests.md], [mocking.md], [refactoring.md], [deep-modules.md], [interface-design.md]) placed at their point of use. Not 5 because none of the five referenced sidecar files exist in the bundle (no references/ or sibling files present), so navigation dead-ends; the references themselves are well organized and only one level deep, keeping this above 3. | 4 / 5 |
Total | 17 / 20 Passed |