Content
92%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 strong, mature skill body: fully executable commands and templates, a gated five-step workflow with real validation loops, and clean one-level-deep reference structure verified against the actual bundle. The only weakness is token budget — roughly 290 lines with a trimmable standards bibliography and occasional non-operational prose.
Suggestions
Move the closing standards bibliography (AICPA guide, COSO framework, ISO/IEC 27002, FedRAMP Consolidated Rules citations) into the per-framework reference files, keeping only the framework-to-reference mapping table in SKILL.md.
Trim non-operational flavor sentences (e.g., "Most readiness work that goes wrong went wrong here") and merge the two handoff tables' surrounding prose to cut the body closer to a lean overview.
Consider moving the full report markdown template into a reference file, keeping inline only the section list and the rule that the JSON drives the prose.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and almost entirely skill-specific knowledge (sample-size tables, test-method hierarchy, fingerprint grammar, evidence-handling rules) rather than concepts Claude already knows, but there is trimmable material: the closing standards bibliography (COSO, AICPA guide, ISO/IEC citations) largely duplicates content carried by the reference files, and a few flavor sentences ("Most readiness work that goes wrong went wrong here") are not operational. This sits at efficient-with-minor-trimming rather than every-token-earns-its-place. | 4 / 5 |
Actionability | Guidance is copy-paste executable throughout: exact read-only commands with portable fallbacks ("gh api repos/<owner>/<repo>/branches/main/protection", "rg -n \"required_approving_review_count\\|enforce_admins\" -g '*.tf'"; portable grep variant), concrete pass conditions ("approver ≠ author, approval before merge, CI green at merge"), worked fingerprint examples, field-by-field tables, a validation command, and a complete renderable report template. | 5 / 5 |
Workflow Clarity | The five steps are clearly sequenced with explicit gates and feedback loops: "Write scope.md before anything else," "run the refutation pass in references/verification.md before anything is confirmed," "Validate it and fix every error" on findings.json, and re-run matching by fingerprint with rules for stale rejected records — matching the top anchor's validation-plus-error-recovery pattern. | 5 / 5 |
Progressive Disclosure | SKILL.md stays at the overview/workflow level while framework detail is split into one-level-deep, clearly linked references (references/soc2.md, references/iso27001.md, references/nist-800-53.md) plus shared format files and scripts (findings-format.md, verification.md, severity-calibration.md, validate-findings.mjs, to-sarif.mjs) — all of which exist in the bundle, with none nested deeper than one level. | 5 / 5 |
Total | 19 / 20 Passed |