skill-improve

Automate quality assurance for AI agent skills by running a test-fix-retest loop that targets static code analysis and rubric-based category compliance. This validates structural integrity and functional requirements before applying persistent changes to the repository.

The best way to maintain and refactor AI agent skills is through an automated refinement lifecycle. This process diagnoses failures, proposes targeted fixes, and ensures operational consistency across the skill library.

Yes, you can roll back changes. The workflow implements a safety-first revert mechanism that allows users to roll back changes if the improvement loop does not yield a better score.

The test-fix loop diagnoses and repairs skill logic by running static and category-based validation checks to identify gaps. It then proposes targeted fixes for identified failures and warnings to ensure skills meet strict quality standards.

No dependencies are required to run automated skill maintenance. The workflow operates independently to validate structural integrity and functional requirements before applying persistent changes to the repository.

AI agent skills fail category compliance checks when gaps in skill logic or structure are detected by rubric-based validation. The automated testing process identifies these failures and proposes targeted fixes to ensure operational consistency.