@its-thepoe/write-a-skill

Agent Skill — author or upgrade Agent Skills: correct SKILL.md frontmatter, triggers, structure, optional reference.md / references/, and clear guidance on skill vs always-on rules (.cursor/rules, CLAUDE.md, AGENTS.md).

What it does

Walks context (audience, greenfield vs recurring workflow), avoids duplication with existing skills, enforces concise SKILL.md + linked deep docs, and checks distribution paths (Cursor, Claude Code, OpenCode, Windsurf, etc.). Helps you ship skills that agents actually invoke when users say the right things.

Why it’s useful

• Reduces broken skills (vague description, no triggers, giant unmaintainable SKILL.md).
• Clarifies when something should be a rule (“always follow”) vs a skill (“when this task appears”).
• Points non-repo users at prompt packs and optional LOVABLE.md when relevant.

Use when

• You’re creating, refining, or scaffolding a skill, slash command, or prompt pack.
• You ask write-a-skill, agent skill structure, or frontmatter for skills.

Install into your agents

npx @its-thepoe/skills@latest install write-a-skill

Contents

• SKILL.md — authoring workflow
• reference.md — templates, checklists, Claude Code / multi-product notes

Docs

• Agent Skills (standard)
• Repository

License

MIT