skill-authoringlisted

# Skill Authoring Create, edit, and evaluate skills in this collection. Every skill conforms to the anatomy in `references/skill-anatomy.md` and passes `scripts/validate-skills.sh`. This skill is the authoring half — `using-skills` routes work to existing skills; this one makes new ones or fixes existing ones. ## When to use - Creating a new skill (new SDLC phase gap, repeated task worth a skill) - Editing an existing skill's frontmatter, steps, or references - Tuning a skill's description triggers (false positives, false negatives, collision with another skill) - Evaluating a skill — does it fire when it should, stay distinct from neighbors, change agent behavior? - Triggers on "write a skill", "edit skill", "skill eval", "tune triggers", "写技能", "改技能", "技能评估" **Not for:** routing a task to an existing skill (use `using-skills`); designing an MCP server (out of scope, see official mcp-builder docs); writing project docs like PRD/README (use `spec`/`oss-polish`). ## Steps ### 1. Check the anatomy before writing Load [${CLAUDE_PLUGIN_ROOT}/references/skill-anatomy.md](${CLAUDE_PLUGIN_ROOT}/references/skill-anatomy.md) — the folder layout, frontmatter rules (name + description + optional disable-model-invocation only), the four body sections (When to use / Steps / Verify / References), the `**Output:**` convention, progressive disclosure (15-150 line SKILL.md, depth in `references/`), and the description hybrid format (`Use when…` + `Triggers on…` + exclusions, ≤1024 char