claude-code-prompts

Why we included this project

Building your own coding agent instead of adopting a turnkey assistant involves plenty of trial and error, and this collection removes some of that guesswork. It studies how Claude Code structures its behavior, then offers independently written templates for the pieces: identity and safety rules, per-tool routing for shell, file edits, and web access, delegation to specialized subagents, and memory for long sessions. You can take a complete system prompt or coordinator prompt, adjust it for your stack and risk policy, and assemble a working agent fairly quickly. The pattern analyses explain why each structure exists, which helps when your agent misbehaves and you are trying to figure out why. There are also ready-to-use Cursor skills if you want integration without wiring everything together yourself.