Concepts

A handful of cross-cutting ideas show up across every skill in this collection. They aren’t part of any single skill’s body, but understanding them makes the skills compose cleanly.

  • Canonical-home pattern: keep one ~/.agents/skills/ directory and let every AI tool symlink into it. One git pull updates every tool from a single source.
  • Dependency tiers: every skill declares its dependencies in three tiers (required, strongly recommended, optional adapter). Don’t conflate them.
  • Severity ladder: High / Medium / Low (with optional Critical for production-blockers) is the shared vocabulary across review and audit outputs in this collection. Tagging is what makes triage possible.
  • Composition over absorption: skills stay strictly in their lane and name neighbor skills for adjacent concerns. You get a small, sharp library where each skill does one thing well.

These are the principles behind why the skills look the way they do. If you’re authoring a new skill for the collection, this is the design lens.

Table of contents