skill-evaluator-catalog-builder

To catalog and evaluate agent skills, you scan local folders or GitHub repositories containing SKILL.md files to extract frontmatter, directory structure, and script languages into a structured JSON catalog. The process performs batch scoring on standardized dimensions to support auditing and comparison.

A skill-catalog.json file contains per-skill entries with extracted frontmatter, script languages, and evaluation scores, alongside aggregate summary statistics. It provides a machine-readable format for easy filtering and benchmarking across your agent skill libraries.

Yes, you can scan GitHub repositories to extract frontmatter from SKILL.md files. The scanning process also extracts directory structure, script languages, license information, and external indicators to build a comprehensive evaluation catalog.

Agent skills are evaluated on standardized dimensions including usage value, security risk, executability, and invocability. These mechanical extraction and batch scoring results merge into a machine-readable catalog for easy filtering and benchmarking.

Yes, the skill catalog builder scales to large agent skill sets by performing mechanical extraction and batch scoring across local folders or GitHub repositories. It processes each SKILL.md file to produce comprehensive per-skill entries and aggregate statistics without manual intervention.