skill-improver

Improving skill instructions involves analyzing gaps between expected and actual AI outputs to classify root causes like missing context or vague scope. You then apply targeted edits to the SKILL.md file and related assets to raise output quality.

Gap taxonomy is a classification framework used during root cause analysis to categorize why AI outputs miss expectations. It identifies issues like instruction insufficiency, instruction excess, context misunderstanding, template deficiency, and procedure order problems.

You update SKILL.md templates by collecting feedback on recent output discrepancies or past usage, classifying the root cause via gap taxonomy, and generating structured change proposals. An improvement changelog records the applied template updates.

Yes, root cause analysis proposes targeted edits rather than full rewrites, preserving existing instruction logic. It generates structured change proposals and records an improvement changelog to track modifications made to the SKILL.md and related assets.

Apply a quality feedback loop when users report poor results, want skill steps reviewed, or revisit past usage to raise quality across similar cases. It requires loading the target skill definition to collect gap details and classify root causes interactively.

You classify AI output gaps by applying gap taxonomy to collected feedback. This taxonomy distinguishes between instruction insufficiency, instruction excess, context misunderstanding, template deficiency, and procedural order issues to pinpoint where to apply fixes.