# skill-optimizer Audits and rewrites an agent skill (SKILL.md) until it passes a measurable bar: 15 analytical passes, a 20-query trigger eval, cross-skill collision checks, and peer routing-mesh seeding, inside a convergence loop. Skills fail in one dominant way: they mis-trigger. `skill-optimizer` treats a SKILL.md as a routing surface first and prose second. Fifteen passes (A–O) run in parallel bundles — correctness, consistency, clarity, frontmatter audit, description rewrite, length budget, anti-AI-ism sweep, whitespace hygiene — and the two that matter most: a **20-query trigger-accuracy eval** (10 should-trigger, 10 near-miss negatives; bar is 9/10 and ≤1/10 false positives) and **cross-skill collision detection** against the installed peer set. Its Pass O is the only pass allowed to edit *other* skills: it seeds bounded, additive deferral edges into peers so whichever skill a lookup surfaces first hands off to the right one — the routing mesh, not just the node. Every run snapshots before writing, fixes all Medium+ findings, re-audits blind with a fresh-context reviewer, and reports a convergence table with per-iteration severity counts. **Use it for:** "optimize this skill", "my skill never triggers", "audit this SKILL.md", new-skill hardening before publishing, post-move structural fixes (`--meta`). **Not for:** an llms file ([llms-deep-optimizer](/skills/llms-deep-optimizer/)) · a prompt in code ([prompt-deep-optimizer](/skills/prompt-deep-optimizer/)) · a one-off prompt (prompt-helper) · rebalancing a whole skill tree ([skill-tree-architect](/skills/skill-tree-architect/)).