skill-improve

Iterative improvement applies static checks to establish a baseline, proposes targeted fixes, rewrites the local skill definition, and re-tests to decide whether to keep or revert based on score shifts.

After re-testing, the workflow compares recorded before and after scores from static and category checks, keeping rewritten changes only if scores improve and reverting automatically if they drop.

Yes, the improvement workflow supports optional category baselines alongside standard static checks to evaluate your skill against category-specific criteria before and after applying targeted fixes.

The best approach is running an automated workflow that generates targeted fixes for gaps found in static checks, rewrites the skill definition, and verifies the impact through re-testing before finalizing.

The iterative loop reverts changes when re-testing shows negative or neutral score shifts, ensuring only modifications that demonstrably improve static and category check baselines are kept.