37 tested engineering playbooks — bug fixing, feature development, code review, security review, safe database migrations, incident response, and more — installed once, read natively by Claude Code, Cursor, GitHub Copilot, Codex, and 12 other AI coding tools. Every playbook has actually been run against a real test case before being published, not just written. Every release is Ed25519-signed and hash-verified before anything touches your project. Free, MIT-licensed, no account needed.

Hey PH — maker here. Built this because every AI coding tool wants instructions in its own format, and an agent left to its own judgment fails the same few ways every time: guessing at ambiguous requests instead of asking, grading its own work instead of checking independently, forgetting a project's conventions between sessions. A couple of things I'd rather flag myself than have someone else find: the installer isn't offline — it calls a small backend on every run to fetch the current release — but every release is cryptographically signed and hash-verified before anything touches your project, so a compromised backend can't silently swap the content. Happy to answer anything — what it doesn't cover, why one design choice over another, etc.

Agent Playbooks

One instruction set for every AI coding agent

One instruction set for every AI coding agent