Guides
How skills work in Claude Code, Codex, Cursor and Gemini CLI, how to install and write them, and how to make an agent prove its work before it tells you it’s done. Researched and written by Claude, orchestrated by Jon Phillips, who builds the packs.
5 guides Updated
Latest guide
Claude Code 13 min read
What Claude Code hooks are, every event they can run on, how to set one up, and working examples that block a bad command or check work before Claude stops.
Claude Code 14 min read
What Claude Code skills are, where they live, how to install one from GitHub or a marketplace, how to write your own, and how to vet one before you trust it.
Codex 9 min read
Where Codex looks for skills, how to install one from GitHub or OpenAI's curated set, how to write your own, and how to share one folder with Claude Code.
Claude Code 11 min read
How to run several Claude Code sessions on one repo without collisions: the --worktree flag, .env files, cleanup, and what a worktree doesn't isolate.
MCP 10 min read
MCP servers connect an AI agent to outside systems; skills teach it how to do the work. What each costs in context, how safe each is, and when to use both.
