skill-loomerylisted

# Skill Loomery A standard workflow for turning expertise into a reusable, reliably-triggering, verifiable **Agent Skill**. Follow this when the user wants to **create a new skill, improve an existing skill's description or instructions, set up evals, or bundle scripts**. ## Core workflow (in order) 1. **Extract real expertise** (see references/best-practices.md, "Start from real expertise") Do not let an LLM invent from general knowledge. Distill from real tasks, project artifacts, and execution traces. 2. **Scope the unit**: one skill should encapsulate one coherent unit of work — too narrow forces multiple skills to load; too broad is hard to trigger precisely. 3. **Scaffold**: `<name>/` directory + `SKILL.md` (required) + optional `scripts/`, `references/`, `assets/`. Copy `assets/skill-template.md` to `<name>/SKILL.md` as a starting skeleton, then edit it. 4. **Write frontmatter** (see references/specification.md) - `name`: ≤64 chars, lowercase/numbers/hyphens only, no leading/trailing/consecutive hyphen, **must match the parent directory name**. - `description`: ≤1024 chars, states both what it does and when to use it, with trigger keywords. 5. **Write the body** (see references/best-practices.md) Step-by-step instructions + a gotchas list + templates + checklists + validation loops. 6. **Optimize the description** (see references/optimizing-descriptions.md) Write imperatively, focus on user intent; iterate with trigger-rate evals over a train/val