skill-improve

You can automate the test-fix-retest loop by running static checks to surface gaps, applying targeted fixes, rewriting the skill, and re-testing to automatically keep or revert changes based on score differences.

Running automated static checks surfaces specific implementation gaps in a skill, enabling you to apply targeted fixes before rewriting and re-testing the skill for evaluation.

You validate a rewritten skill by re-testing it and comparing the new score against the previous baseline, automatically reverting the changes if the score drops and keeping them if it improves.

Yes, parsing the skill's SKILL.md frontmatter is required to execute the automated improvement cycle, as it integrates with the test command and enables optional category-aware checks for deeper evaluation.

Yes, you can use optional category-aware evaluation to ensure your skill aligns with a broader rubric during the test-fix-retest loop, providing deeper validation beyond standard static checks.