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 highly actionable, well-validated skill body with excellent workflow sequencing and a sound one-level-deep reference structure. Its main weakness is length: several sections duplicate flag-level detail already delegated to references, and exit-code guidance appears in three places.
Suggestions
Trim the `flags --retirement` and CODEOWNERS health paragraphs to one representative example plus a pointer to references/patterns.md, keeping the trigger conditions and the advisory nature in the body.
Consolidate exit-code guidance into the single Exit Codes section and have Agent Rule 2 reference it, removing the repeated inline explanations in the security and audit sections.
Shorten Agent Rules 14 and 15 to their decision rules and move the step-by-step snapshot/inspection mechanics into references/similar-code.md and references/mcp.md, which they already point to.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and tool-specific with almost no generic-concept padding, but several sections enumerate flag-level detail the reference files already carry — e.g., the `flags --retirement` paragraph lists every reason code, and Agent Rules 14–15 are multi-sentence walls. Exit-code guidance is repeated across Agent Rule 2, the Exit Codes section, and inline in security/audit workflows, so noticeable tightening is possible. | 3 / 5 |
Actionability | Fully executable throughout: dozens of copy-paste-ready commands with exact flags, JSON output shapes, a worked judgment-file schema for the review walkthrough, exit-code semantics, and non-TTY `--yes` requirements. Specific examples cover the common cases (dead-code, audit, dupes, fix cycle, workspaces, baselines). | 5 / 5 |
Workflow Clarity | Multi-step processes are explicitly sequenced with validation checkpoints and error-recovery loops: the destructive `fix` path is dry-run → review → `fix --yes` → verify; the review walkthrough handles stale-hash rejection and re-fetch; baselines and CI gates have save/gate sequences. Feedback loops are explicit rather than implied. | 5 / 5 |
Progressive Disclosure | Good structure: seven real one-level-deep reference files (cli-reference, mcp, issue-types, gotchas, patterns, similar-code, node-bindings), each listed with a description in a References section, with bulk catalogues properly delegated. Minor gap: some flag-level detail (flags retirement reasons, CODEOWNERS health scoring fields) remains inline in the body where a reference file would be the right home. | 4 / 5 |
Total | 17 / 20 Passed |