Content
63%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-sequenced, largely actionable skill with concrete commands and a clear scoring methodology, weakened by monolithic structure (no bundle files despite ~350 lines of content), repeated external quotations, and pseudocode dependencies on undefined helpers and an unshipped script. The two lowest dimensions, conciseness and progressive disclosure, share a root cause: material that should be offloaded to reference files is duplicated inline.
Suggestions
Split the coverage debt ledger, the per-format parsing details (Step 2), and the References quotations into one-level-deep reference files (e.g. references/debt-ledger.md, references/parsers.md), leaving SKILL.md as a lean overview with well-signaled pointers.
De-duplicate the external quotations: the test-pyramid and McCabe quotes each appear twice in the body and again in References — cite them once where they justify a design decision.
Either ship scripts/coverage_targets.py (the CI snippet depends on it) or replace the invocation with the inline commands it would wrap, and define or inline the normalize/mccabe_complexity/git_churn helpers so the scoring code is executable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and skill-specific, but the external quotes (test pyramid, McCabe thresholds) appear twice in the body and again verbatim in References, and the ~30-line rendered example plus the long anti-pattern table add padding. Mostly efficient, but noticeably more trimming available than the 'minor instances' of anchor 4. | 3 / 5 |
Actionability | Concrete executable commands (radon, complexity-report, git churn, CI YAML) and explicit scoring weights with formulas. Not a 5 because risk_weight calls undefined helpers (mccabe_complexity, normalize, git_churn) and the CI snippet runs scripts/coverage_targets.py, which is not shipped in the bundle. | 4 / 5 |
Workflow Clarity | Steps 1-8 are clearly sequenced with preconditions flagged (fetch-depth: 0 for churn history, skipping pragma-ignore lines). The workflow is advisory/read-only so the destructive-batch cap does not apply, but explicit validation checkpoints (e.g., verifying parser output) are absent, keeping it below anchor 5. | 4 / 5 |
Progressive Disclosure | There is no bundle at all (no references/, scripts/, or assets/ directories), so the entire ~350-line payload is inlined in SKILL.md. Section headers are well-organized, but the debt-ledger section, per-format parsing details, and reference material clearly belong in separate one-level-deep files, matching anchor 3. | 3 / 5 |
Total | 14 / 20 Passed |