skill-improve

To automate skill improvement, run a structured test-fix-retest loop that diagnoses gaps, proposes targeted fixes, and revalidates changes. This evaluates skill structure, frontmatter gaps, and rubric alignment to iteratively apply precise modifications.

Static check failures diagnosed during gap analysis include missing frontmatter, insufficient phases, and missing verdict keywords. The workflow evaluates skill structure and rubric alignment to identify these specific gaps before proposing targeted fixes.

Fix missing frontmatter and rubric alignment issues by generating precise edits to SKILL.md and related docs. The workflow proposes targeted modifications based on diagnosed gaps and validates the changes with follow-up tests.

Yes, the skill improvement workflow compares static and category scores before and after changes. It includes safety checks for revert if test scores drop, maintaining a reproducible workflow with an audit trail to decide on keeping or reverting modifications.

Yes, you can use this automated diagnostic workflow for a single skill at a time. It parses arguments to baseline tests, diagnose gaps, propose fixes, and retest until scores improve or remain stable for that specific skill.