skill-improve

To fix static check failures, run a test-fix-retest workflow that identifies exact failing checks, proposes targeted edits to resolve missing elements, and retains changes only if the combined score improves.

A test-fix loop for skill improvement is a deterministic workflow that runs static and category rubric checks, diagnoses specific failures, applies minimal targeted edits, and re-tests to confirm score improvement before keeping or reverting changes.

Category rubric scoring is applied by running optional category-specific tests from the CCGS testing framework after assigning a category to the skill, producing a diagnosis tied to specific check IDs for targeted fixes.

Yes, you can improve a skill without full rewrites by mapping each failing or warning check to a precise missing element and writing only necessary before and after edits after user approval.

You should revert a skill when the combined score from static and category rubric tests does not improve after applying targeted edits, ensuring no regressions are introduced.

No external dependencies are required for skill improvement, as it relies on parsing the skill argument and running built-in static and category tests to produce diagnoses and apply minimal edits.