Thank you for your interest in contributing to KIP! This guide will help you get started.
- Use GitHub Issues for bug reports and feature requests
- Include your Claude Code version and OS
- For capture accuracy issues, include the conversation context that was (or wasn't) captured
- Fork the repository
- Create a feature branch:
git checkout -b feature/my-improvement - Make your changes
- Run evals: ensure all test cases pass
- Submit a Pull Request
- All existing evals pass
- New features include eval test cases
- English-first in user-facing content
- No external dependencies (pure prompt skill)
- Code follows existing style conventions
# Clone
git clone https://github.com/DavidKim0326/KIP.git
cd KIP
# Install as Claude Code plugin (for testing)
# Option 1: Plugin marketplace
/plugin marketplace add DavidKim0326/KIP
/plugin install kip
# Option 2: Manual
cp -r . ~/.claude/skills/kipAdd new test cases to evals/evals.json:
{
"id": 9,
"mode": "CAPTURE",
"prompt": "Your test prompt here",
"expected_output": "Expected behavior description",
"expectations": [
"Specific expectation 1",
"Specific expectation 2"
]
}Each expectation should be independently verifiable.
- New natural language patterns for deferred task detection
- Additional language support beyond English/Korean
- False positive reduction
- Token budget optimization
- Accessibility improvements
- Alternative display formats
- Better 🔥 detection algorithms
- Reduced false positive rate
- New signal types
- Markdown: ATX headings, fenced code blocks with language tags
- JSON: 2-space indentation
- JavaScript (hooks): ES6+, JSDoc for exported functions
By contributing, you agree that your contributions will be licensed under the MIT License.