skill-doctorlisted

# skill-doctor Diagnose one Agent Skill against the gold standard and prescribe fixes. Self-contained: the rubric is carried here as prose, so it runs anywhere with no extra tooling. It **recommends first, then edits the file in place only after you confirm.** The rubric and the ecosystem evidence behind each check live in [`reference/rubric.md`](reference/rubric.md). Read it when you need the detailed scoring or the corpus numbers; the workflow below is enough for a normal review. Open questions and forward-looking ideas that aren't scored yet — noise complaints, model-generation-aware maintenance, and more — live in [`reference/consider-this.md`](reference/consider-this.md). ## When to use - You have a `SKILL.md` and want it audited or hardened. - Someone asks "is this skill any good?" or "make this best practice." ## When NOT to use - Authoring a skill from a blank page → use `skill-creator`. - Grading an entire repo of skills at once → use the skill-map auditor (`audit_repo.py`) or the `skill-audit` GitHub Action. ## Workflow Follow these steps in order. Do not skip the interview — the highest-value fixes (tool scoping, data handling, anti-triggers) cannot be inferred from the file alone. ### 1. Read the skill Read the target `SKILL.md` and its sibling files (`reference/`, `scripts/`, `assets/`). Note: frontmatter keys present, description length, whether the body has a WHEN trigger and an anti-trigger, headings, and any sibling files. ### 2. Static pass (no que