skill-improve

Automating test-fix-retest cycles for debugging custom skills involves running a static baseline test, applying targeted fixes, and re-testing to measure score changes. This validates improvements systematically and reverts ineffective fixes automatically.

Measuring quality assurance improvements in a target skill requires running static tests to establish a baseline, applying fixes, and re-testing to evaluate score changes. The workflow reverts modifications if the score does not improve.

To start an automated skill improvement loop for CI-CD workflows, provide the target skill name as an argument. The process orchestrates argument parsing, baseline checks, diagnosis, fixes, and retesting automatically.

The automated testing and debugging workflow requires a SKILL.md file located at .claude/skills/[name]/SKILL.md. It relies on this specific file structure to perform validations using /skill-test commands.

You can use this automated fix and validation process across skill domains where formal testing workflows exist. It applies targeted fixes and validates changes universally, independent of the specific skill subject matter.

If applied fixes do not improve the skill test score during the re-testing phase, the automated workflow reverts the changes. This ensures only modifications that demonstrably enhance quality assurance are kept.