AI Coding Boilerplate

Agent-friendly project structure for humans and AI coding agents

An open-source, AI-agent-friendly project boilerplate that bakes in deterministic practices so both humans and autonomous agents can work in a sane, repeatable way. The goal isn’t to lock agents down — it’s to give them clear rules, precise context, and hard physical gates.

Structure:

  • .ai/ (The Context Hub): RULES.md (no secrets, no force push), ARCHITECTURE.md (system map), and MEMORY.md (decision logs). Agents are prompted to read these before touching code
  • STATE.md (The Dispatch Board): Current focus, active blockers, and strict Definition of Done checklist — agents and humans always know what “done” means
  • Ironclad Hooks: Pre-commit and commit-msg hooks enforce secrets scanning, debug-artifact checks, test runs, Docker build validation, and Conventional Commits
  • Language-Agnostic: Works identically for Python, Node, Go, etc. Includes Docker sandboxing and a generator script to instantly spin up new agent-ready projects