skill-judge

To evaluate Skill quality, you assess the knowledge delta a SKILL.md delivers against the model's baseline, checking frontmatter validity, description quality, and progressive disclosure to ensure reliable activation and minimal token waste.

Common anti-patterns in Skill design include poor frontmatter validity and missing loading triggers. You can fix them by detecting these issues, supplying NEVER lists with concrete reasoning, and guiding three-layer content loading for on-demand resources.

Progressive disclosure works by structuring a SKILL.md body to load only required content initially, utilizing loading triggers across optional references and asset directories on-demand to ensure activation and minimize token consumption.

Generating a formal Skill evaluation report requires a valid SKILL.md with a concise name and description. The process analyzes frontmatter, description quality, and loading triggers to produce actionable improvements for skill packages.

Yes, you can assess frontmatter validity and loading triggers for skill packages without external dependencies. The Skill operates standalone to quantify value by measuring expert knowledge delivered as knowledge delta against the baseline.