If you use Cursor, GitHub Copilot, or Amazon Q, you know how powerful custom rules can be for guiding AI behavior. But right now, everyone just copies `.cursorrules` files around manually. Once copied, they're orphaned – no updates, no version control, no way to know if changes will break your AI's behavior.

ARM treats AI rules like npm packages. You can:

- Install versioned rulesets: arm install awesome-cursorrules/python - Connect to Git registries like PatrickJS/awesome-cursorrules - Keep rules synced across projects with semantic versioning - Target multiple AI tools (Cursor, Copilot, Amazon Q) with different layouts

The key insight: AI rules should be dependencies, not copy-pasted files. When the awesome-cursorrules repo updates their Python rules, you want to choose when and how to upgrade, just like any other dependency.

Quick example:

---------------------------

# Connect to a registry arm config registry add awesome-cursorrules https://github.com/PatrickJS/awesome-cursorrules --type git

# Install Python rules to Cursor arm config sink add cursor --directories .cursor/rules arm install awesome-cursorrules/python

# Later, update when you're ready arm update awesome-cursorrules/python

----------------------------

It creates arm.json and arm-lock.json files (like package.json/package-lock.json) so your whole team gets the same AI behavior.

One-liner install:

curl -fsSL https://raw.githubusercontent.com/jomadu/ai-rules-manager/ma... | bash

GitHub: https://github.com/jomadu/ai-rules-manager

Would love feedback from other folks dealing with AI rule management chaos!