skill-evaluator

Skill evaluation applies automated validation and manual review criteria to audit SKILL.md artifacts for naming conventions, description quality, content structure, and references. The Skill generates a comprehensive report with dimension scores, actionable recommendations, and a pre-publication checklist to ensure compliance with Anthropic's standards.

The evaluation report scores skills across multiple dimensions—naming, description quality, content density, structure, and anti-patterns—then provides strengths, areas for improvement, and prioritized recommendations. A pre-publication checklist summarizes next steps for refinement before release.

Yes. The Skill runs automated structural validation to check naming conventions and content limits, reducing manual review burden. It supports batch evaluation across skills to ensure consistent quality standards across your publishing pipeline.

You need a completed SKILL.md artifact with title, description, features, and use cases defined. PyYAML is the only dependency required to run the evaluation script against your skill's metadata and content.

The evaluation identifies specific deficiencies in structure, naming, and content quality, then returns concrete recommendations to fix each issue. The report prioritizes improvements by impact to help you address the most critical problems first.

Yes. The Skill evaluation applies to review, audit, and improvement workflows for skills in any state—from draft to pre-publication. It adapts to varying content density and maturity levels while maintaining consistent best-practice benchmarks.