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 body is dense, repo-specific, and immediately actionable: every command is copy-paste ready with flags explained, and the numbered rules name exact tokens and allowed fields so a user can act without inference. Deductions are minor — some guidance is duplicated between Core Concepts and Best Practices, the fix-and-re-run loop is implicit, and the Core Concepts section mixes two kinds of context.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Nearly every line carries repo-specific facts Claude cannot infer ('Codex loads `.agents/skills/`. Cursor loads a smaller `.cursor/skills/` subset', '`validate-skills.js` only checks that `SKILL.md` exists'). The '$CLAUDE_PROJECT_DIR' → '$(pwd)' rewrite advice is duplicated between 'Core Concepts' and 'Best Practices', and the Anti-Patterns lead sentence is awkwardly dense, so minor trimming remains — anchor 4, not 5. | 4 / 5 |
Actionability | Commands are copy-paste ready with every flag shown ('node scripts/ci/validate-skill-host-compat.js --inventory', '--strict-checklist') plus slash-command forms, and the rules name exact tokens ('$CLAUDE_SESSION_ID', '$ARGUMENTS') and exact allowed frontmatter fields. Fully executable guidance covering the common cases — anchor 5. | 5 / 5 |
Workflow Clarity | Usage sequencing is explicit ('Run this script before `npm test` when the change is a new skill') and the four numbered check categories tell the user how to interpret findings. The fix-and-re-run loop is implicit rather than an explicit checkpoint and no sample output is shown, keeping it below anchor 5; not 3 because the lint gate itself is the validation checkpoint and the single action is unambiguous. | 4 / 5 |
Progressive Disclosure | Sections are well-organized (When to Activate, Core Concepts, Code Examples, Anti-Patterns, Best Practices, Related Skills) with no nested references and no content that clearly belongs in a separate file. At ~96 lines with a dense 'Core Concepts' section that mixes repo-layout facts with the four rules, it sits just below the clean single-file ideal of anchor 5. | 4 / 5 |
Total | 17 / 20 Passed |