Content
96%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.
An excellent single-file skill: lean, densely actionable, and sequenced as a clear diagnostic workflow with feedback loops. The only structural gap is that the ~100-line body keeps reference-grade material (the jq filters and schema facts) inlined rather than split into a reference file.
Suggestions
Move the jq filter library and/or the schema-facts list into a references/ file (e.g. references/filters.md), keeping only the workhorse failure-extraction filter inline in SKILL.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body assumes Claude's competence — it never explains what Ginkgo, jq, or Go are — and every line delivers non-obvious knowledge (e.g. the ~9x token cost of the raw JSON, that --json-report does not suppress console output). | 5 / 5 |
Actionability | Guidance is fully executable: a copy-paste ginkgo invocation, five tested jq filters targeting the real schema, concrete signal names (SIGUSR1/SIGINFO), and specific flag combinations covering the common diagnostic cases. | 5 / 5 |
Workflow Clarity | The workflow is explicitly numbered (run once, read the terminal verdict, jq-query if failures, clean up) with a conditional checkpoint, and the "Reproduce, then narrow" section plus the three-traps section form a genuine feedback loop for error recovery. The destructive/batch cap does not apply since diagnosis is read-only. | 5 / 5 |
Progressive Disclosure | Sections are well-organized and external pointers (ginkgo:writing-specs, ginkgo:reporting, docs URL) are one level deep and clearly signaled, but at ~100 lines the jq filter library and schema-facts block are reference-grade material inlined in the main file, with no bundle files used to split them out. | 4 / 5 |
Total | 19 / 20 Passed |