skill-craft

You can audit Claude Skills by applying a structured 8-module quality framework that checks frontmatter metadata, module structure, and reference integrity to produce actionable reports.

Evaluating multiple Claude Skills is done through global audits that run discovery validation and decision-gate checks across all targeted Skills to identify and repair structural anti-patterns.

Automated validation for Claude Skill scripts and references works by enforcing frontmatter metadata, checking module structure, verifying reference integrity, and running validation scripts to generate reports.

Yes, you can create new Claude Skills using a standardized 8-module quality model that enforces completeness principles and prevents common anti-patterns during the initial creation process.

A Claude Skill evaluation checks for 7 specific anti-patterns alongside 3 completeness principles to ensure proper module structure, valid references, and correct frontmatter metadata.

No external dependencies are required; you only need to copy the SKILL.md, references directory, and scripts into your Claude skills directory and run the evaluate command on your target Skill.