Content
86%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 high-quality reference skill body: dense, non-obvious Ginkgo semantics delivered with opinionated guidance and concrete code/CLI examples, well-sectioned with clear cross-references. The only soft spots are elided code bodies and the absence of a 'what to do when the flake hunter catches something' follow-through.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and dense throughout: it spends tokens only on Ginkgo-specific semantics Claude cannot be assumed to know ('runs serial specs last, on process #1', 'BeforeAll/AfterAll are legal only inside an Ordered container', 'you cannot mark an individual spec inside an Ordered container Serial') plus opinionated stances ('retries are explicit technical debt'), and defers sibling topics via pointers. | 5 / 5 |
Actionability | Concrete Go decorator usage and copy-paste CLI commands ('ginkgo --until-it-fails -p --randomize-all', 'ginkgo --flake-attempts=3'), but several example bodies are elided stubs ('func() { ... }', 'AfterAll(func() { ... })') — minor gaps versus fully copy-paste-ready code. | 4 / 5 |
Workflow Clarity | Decision paths are clearly sequenced (expose flakes first, tolerate 'judiciously, as a last resort'; 'Prefer one It + By by default — Reach for Ordered when...'), and no validation cap applies since nothing destructive is performed. It falls short of a 5 because there is no error-recovery/feedback-loop structure (e.g., what to do once --until-it-fails catches a break). | 4 / 5 |
Progressive Disclosure | No bundle files exist, and the body is appropriately self-contained: well-organized sections with clearly signaled one-level-deep cross-skill pointers ('→ ginkgo:decorators', '→ ginkgo:ci') and a closing Related section. Nothing is buried or inlined that belongs in a separate file. | 5 / 5 |
Total | 18 / 20 Passed |