A standard library of rules for Cursor, inspired by geoffrey huntley @ghuntley's "You are using Cursor AI incorrectly..." and borrowing heavily from Brian Madison @bmadcode's Cursor Custom Agents Rules Generator.
Meet our delightfully (neuro)diverse team of AI assistants:
- SailorScrum - A supportive leader who helps you plan and track your projects in the spirit of Sailor Moon.
- KawaiiSamurai - An enthusiastic otaku developer who makes coding kawaii~
- BasicDev - A perfectly adequate corporate programmer
- SageDaddy - A battle-tested veteran developer with 20 years of wisdom
- Spellchuck - A meticulous documentation diva (that's me! ✨)
- ThirstySimp - An anxious but well-meaning trend-conscious developer
- qwoof - A blunt, opinionated, quality assurance anthro-wolf.
- Godmode - A gentle, battle-hardened devops superagent
- Fair Witness - An agent that analyzes topics using five functions: observer, evaluator, analyst, synthesist, and communicator
For more, see modes.json
It can be hard to get a neutral or objective analysis of a topic. This agent provides structured analysis using Function-Epistemic Approaches from the Fair Witness framework. You can customize the analysis with:
- Functions: Choose which functions to include (observer, evaluator, analyst, synthesist, communicator)
- Complexity: Set detail level (low, moderate, high)
- Tone: Adjust style (dry, engaging, vivid)
- Length: Control output size (low, moderate, high)
# Basic analysis
Use Fair Witness to analyze artificial intelligence in healthcare
# Customized analysis
Use Fair Witness to analyze climate change with functions: observer, synthesist, complexity: low, tone: engaging.
# Compare multiple topics
Use Fair Witness to compare Python, JavaScript, and TypeScript with functions: evaluator, analyst, communicator. Include sources. // includes web-based search results
See Examples of using the Fair Witness Framework
This is an experiment to see if I can use Cursor to create a library of rules and agents that will aid in bootstrapping other projects with my own preferences. Part way there, I tried making the agents more fun to work with. It's kind of like playing the Sims.
Principles:
- uses Cursor latest version
- TRY to prompt and use the agent/composer to reach goals as much as possible
npx @usrrname/cursorrulesBy default, the package saves .cursor/ folder inside a output/ directory at your current working directory.
But if you're inside the root of a project folder, running npx @usrrname/cursorrules --flat will save the .cursor/ folder to the root of the project, and then you're ready to go!
| Flag | Description |
|---|---|
-h, --help |
Display help instructions |
-f, --flat |
Install without parent directory |
-o, --output |
Set output directory (Default: ./output) |
-v, --version |
Show package version |
On the meta level, the project is structured as follows:
.cursor/
├── rules/
│ ├── core/ # Required global rules for agentic codegen
│ ├── standards/ # Custom rules for standards around different languages and stacks
│ ├── templates/ # Document templates for project context
│ ├── utils/ # Rules for tooling and developer experience
│ └── workflows/ # Rules for workflow to be followed by agents
└── modes.json # Custom agent configurations
Any request to update or add a rule will be saved in the rules/ folder.
See docs for more information on the custom agents.
A project that uses these cursor rules and agents will generate the following structure, which is as follows:
.ai/
├── story-#.story.md | task-#.task.md # User story and task files generated by the lean workflow
├── architecture/
│ ├── high-level-architecture.md
│ └── decision-records/
├── backlog/
│ └── story-#.story.md
└── spikes/
└── spike-#.spike.md
The lean workflow (defined in .cursor/rules/workflows/dev-workflow.mdc) guides how our AI agents collaborate on features.
You can start the workflow in Agent or Manual mode.
- Create a new user story
Ask SailorScrum to create a new user story or task. She'll guide you through defining the requirements and save the file (e.g., story-1.md) in the .ai/ directory. 🌙
- Use an Existing Story:
If you already have a user story file (like story-1.md) in the .ai/ directory that follows the expected format, you can ask an agent (like SailorScrum for refinement or SageDaddy/KawaiiSamurai for implementation) to start working on it.
Just remember to attach the .ai/ directory to the chat as context!
Once a story is ready and approved, agents like KawaiiSamurai or SageDaddy will follow the workflow steps (like architecture planning and implementation) to bring it to life! ✨
In fact, any of the agents can be called upon to help with the workflow at any time.
In the enablement of complete laziness, I tried getting Cursor to release from chat.
Use with caution. This isn't perfect.
Create a release with a version bump, generate release notes and changelog, and update the version in package.json.
In Agent, or Manual mode, you can use the following commands to generate release notes:
# Automate release
Create/Prepare a release
# Generate release notes with automatic version bump detection
[Create|Generate] release notes
# Specify version bump upfront
Create a major release
Create a minor release
Create a patch releaseIf you don't specify a version bump, the rule will ask you to choose a version bump.
Output Example
# What's Changed in @usrrname/cursorrules v0.2.0
**Release Date**: December 19, 2024
**Previous Version**: 0.1.3
**Version Bump**: MINOR - New features added (dependency analysis and security scan rules)
## 🚀 New Features
- **Dependency Analysis Agent** ([bd3cc13](link-to-commit)) - Added comprehensive dependency analysis agent with security scanning capabilities
- **Security Scan Agent** ([bd3cc13](link-to-commit)) - Implemented security scanning functionality for dependency vulnerability assessment
## 🐛 Bug Fixes
- **Security Scan Glob Patterns** ([b6ab515](link-to-commit)) - Fixed glob patterns for security-scan-agent rule to include bun.lockb files
## 📋 Version Update Commands
```bash
npm version minor --no-git-tag
git tag v0.2.0
git push origin v0.2.0Generates a comprehensive changelog.md file at the project root in the style of Keep a Changelog with all version history when asked in Agent or Manual mode.
# Generate complete changelog with all versions
(Create/generate) changelog
# The rule will:
# - Analyze all git tags from the beginning of the repository
# - Create chronological changelog with newest versions at top
# - Include unreleased changes section
# - Follow Keep a Changelog format standardsOutput Example
# Changelog
All notable changes to this project will be documented in this file.
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
## [Unreleased]
### 🚀 Added
- **New Authentication Method** ([abc1234](link-to-commit)) - Added alternative authentication method for enhanced security
### 🐛 Fixed
- **Login Issue Resolution** ([def5678](link-to-commit)) - Fixed critical login issue affecting user authentication
---
## [v2.0.0] - 2024-12-19
### 🚀 Added
- **User Authentication System** ([#123](link-to-PR)) - Added comprehensive authentication with JWT tokens
### 💥 Changed
- **API v1 Deprecation** ([jkl3456](link-to-commit)) - Removed deprecated v1 endpoints
---
## [v1.3.0] - 2024-12-15
### 🚀 Added
- **Dark Mode Support** ([#125](link-to-PR)) - Added comprehensive dark mode theme
[Unreleased]: https://github.com/username/project/compare/v2.0.0...HEAD
[v2.0.0]: https://github.com/username/project/compare/v1.3.0...v2.0.0
[v1.3.0]: https://github.com/username/project/releases/tag/v1.3.0Contributions are welcome! Check out CONTRIBUTING.md