skill-improve

To improve an existing skill, you run baseline static tests to identify failures, diagnose specific gaps, apply targeted revisions, and perform measured retesting to verify whether the changes actually improve quality.

The best way to diagnose skill quality assurance gaps is mapping test results to focused changes, applying targeted revisions with user approval, and comparing before-and-after scores to keep only effective improvements.

Yes, you can rollback skill revisions. The process supports conditional rollback by comparing before-and-after retesting scores and reverting any changes that do not measurably improve the target skill file.

You run category validation checks by applying category rubrics during baseline testing. This records specific warnings and gaps, ensuring the skill meets category-specific requirements before applying revisions.

You need access to the target skill file, the skill-test command, the testing framework catalog, and read and write permissions to apply user-approved file updates during the skill maintenance workflow.

Use automated skill improvement when a skill lacks safety prompts, structured phases, or clear follow-up guidance, ensuring you preserve sections that already pass validation while fixing diagnosed gaps.