Cursor AI Coding Playbook
Lightweight rules to help AI copilots write better code, reduce compute cost, and keep teams moving fast.
This repo contains the Cursor Rules I use across projects to ensure high-quality output, predictable model behavior, and efficient use of compute. These rules guide architecture, planning, code style, and model-choice discipline when working with AI copilots.
They are intentionally small, opinionated, and built for speed.
⸻
📘 What’s Inside
Each rule lives as a modular .mdc file inside rules/. Cursor automatically loads and applies them in projects using this repo via worktrees.
- Architecture Principles
High-leverage architectural constraints for small and mid-sized AI-driven products. Helps copilots generate stable, scalable solutions without over-engineering.
- Clean, Simple, DRY
Forces copilots to favor clarity, simplicity, and reuse. Reduces surface area, makes diffs smaller, and keeps projects maintainable.
- Full-Stack Programming
Guidance for end-to-end implementation—React, backend, DB—so AI agents produce coherent, integrated features rather than isolated fragments.
- Effective Planning
Planning templates and conventions that keep tasks atomic, scoped, and easy to execute. Prevents runaway context, multi-feature merges, and diff bloat.
- Model Economy
Rules for using Sonnet, Codex, and Cursor efficiently—minimizing token burn while maintaining speed. Includes: • model-check header • CHEAP / STANDARD / PREMIUM task classification • guardrails against slow or expensive actions
- Models Rule (Meta-Rule)
Lightweight behavioral constraints for Cursor’s model itself (e.g., output format, cost-aware classification). Ensures consistent model behavior across tasks.
⸻
🎯 Why This Exists
AI copilots are powerful, but without guardrails they can: • escalate models unnecessarily • rewrite full files • load huge contexts • thrash between tasks • burn compute without improving results
These rules solve that by giving the models a shared language, shared constraints, and consistent execution framework.
⸻
🛠 How to Use
You can consume these rules in three ways:
-
Add as a Git worktree (recommended)
git worktree add .cursor cursor-worktree
Then use Cursor’s “Project Rules” interface to enable them.
-
Clone the repo and copy the rules folder
Simple, direct, portable.
- Import specific rules selectively
Each .mdc file is standalone and safe to use independently.
⸻
📄 Repo Structure
/rules architecture-principles.mdc clean-simple-dry.mdc full-stack.mdc implement-plan.mdc model-economy.mdc models.mdc
🤝 Contributing
This playbook will evolve as Cursor, Sonnet, and Codex evolve. PRs and discussions are welcome—especially around: • improved model-economy heuristics • React/Next.js patterns • better planning formats • safety or correctness guardrails
