A good .cursorrules file is the difference between AI code that fits your project and AI code that fights it. I refined mine over 8 months. Here is the template and the reasoning behind each section.
What should every .cursorrules file contain?
Three sections. Project context: what stack, what architecture, what conventions. Constraints: what the AI should never do (hardcode secrets, modify config files, add new dependencies without approval). Instructions: how the AI should behave (ask before writing to disk, show diffs, test after every change).
How do you handle multiple .cursorrules for different project areas?
I use project-level .cursorrules in subdirectories. /frontend/.cursorrules has frontend-specific rules. /backend/.cursorrules has backend rules. Cursor reads the nearest .cursorrules file. This lets me give detailed constraints to the backend AI (no SQL injection, parameterized queries required) without cluttering the frontend instructions.
Leave a Reply