This repository provides custom instructions for the Cursor IDE, specifically optimized for its AI agent capabilities. It aims to enhance the agent's analytical skills, enable autonomous processing, and prevent common LLM coding issues like version rollbacks, duplicate resource generation, unintended design changes, and infinite error loops. The target audience includes Cursor users seeking to improve their AI-assisted coding workflows.

How It Works

Version 5 of cursorrules is optimized for Cursor Agent's autonomous processing, leveraging Claude 3.5 Sonnet for enhanced analysis. It structures custom instructions into separate files (e.g., technologystack.md, directorystructure.md) referenced via the @ symbol within the main global.mdc file. This modular approach allows for greater flexibility in managing and updating technical stacks and directory structures, facilitating easier customization for specific projects and environments.

Quick Start & Requirements

1. Add a new rule in Cursor's "Rules" tab, naming it (e.g., my-project).
2. Copy global.mdc from this repo into the created .cursorrules folder.
3. Delete the file created in step 2 to allow Cursor to create the folder.
4. Save technologystack.md and directorystructure.md to your project's root directory (or skip if not needed, removing corresponding sections from global.mdc).

Highlighted Details

• Optimized for Cursor Agent's autonomous processing.
• Addresses LLM coding issues: version rollbacks, duplicate generation, design changes, error loops.
• Modular design using @ file references for flexibility.
• Tested with Claude 3.5 Sonnet and ChatGPT (o1 pro).

Maintenance & Community

• No official support, but feedback is welcomed.
• Author shares Cursor-related information on X (Twitter).

Licensing & Compatibility

• MIT License.
• Compatible with Cursor IDE's Project Rules feature.

Limitations & Caveats

Conflicting instructions or excessive content within "Rules for AI" may reduce effectiveness. Validation is recommended for specific environments and use cases, as results may vary.