Content
82%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.
The skill body is dense, actionable guidance with a concrete grep sweep and a clear rules-to-sweep-to-sync workflow. Its main flaws are the thrice-stated PRD exception, the pair-check script being referenced but not given as an executable step, and one misplaced bullet that confuses the research-docs section.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and lean overall, and the smell/why/fix table earns every token, but the PRD exception is stated three times ('History, corrections, and discovery narratives belong in PRDs', the 'This is a different rule from prd-writing' paragraph, and 'This does not apply to PRDs'), which is trimmable repetition. | 4 / 5 |
Actionability | The sweep commands are copy-paste ready ('grep -rniE "as of 202|caught in review|..."' and 'grep -rn "—"') and the smell table pairs each violation pattern with a concrete fix, covering the common cases for an instruction-only skill. | 5 / 5 |
Workflow Clarity | A clear sequence is present (apply rules, run the sweep, 'Read every hit in context', fix violations, sync the three trees) with the sweep as an explicit validation checkpoint, but there is no re-run/feedback loop after fixes and 'scripts/check_skill_pairs.py' is described rather than given as a command to run. | 4 / 5 |
Progressive Disclosure | A well-sectioned single file with no unnecessary external references, which is appropriate at this size; however, the bullet 'Research docs don't need cross-tree sync ... so this section doesn't apply to them' sits inside the 'Also applies to: research/audit docs' section, contradicting the section it is placed in. | 4 / 5 |
Total | 17 / 20 Passed |