skill-judge

Quick Review

Exceptional meta-skill that encodes expert knowledge for evaluating AI skills. Provides a sophisticated 8-dimensional evaluation framework (120 points) with deep insights into knowledge delta, mindset patterns, and anti-patterns that Claude would not generate independently. The description is comprehensive with clear WHAT (multi-dimensional scoring), WHEN (reviewing/auditing SKILL.md files), and keywords (skill evaluation, SKILL.md). Structure is excellent with clear protocol, decision frameworks, and quick reference checklist. The novelty is outstanding—this codifies nuanced judgments about token economics, knowledge externalization, and the subtle distinction between expert/activation/redundant knowledge that would require significant reasoning tokens for a CLI agent to derive. Minor improvement: description could add more trigger keywords like 'score', 'audit', 'quality check'. This is production-ready expert-level content that demonstrates the very principles it teaches.