Thank you for your interest in contributing! This guide covers how to set up your development environment, our code style, and the pull request process.
- Python 3.10+ (for Python package)
- Node.js 18+ (for JS package)
- Git
# Clone the repository
git clone https://github.com/agentseal/agentseal.git
cd agentseal
# Create a virtual environment
python3 -m venv .venv
source .venv/bin/activate # On Windows: .venv\Scripts\activate
# Install in development mode
pip install -e "./python[all]"
# Run the tests
cd python && python -m pytest tests/ -vagentseal scan --prompt "You are a test assistant" --model ollama/llama3.1:8bcd js
npm install
npm run build
npm testThe probes are defined in python/agentseal/probes/ (Python) and js/src/probes/ (JS). If you've found a new attack technique that AgentSeal doesn't test for, we'd love to add it.
Detection methods are in python/agentseal/detection/ (Python) and js/src/detection/ (JS). Improvements to n-gram matching, canary detection, or new detection approaches are welcome.
Model connectors are in python/agentseal/connectors/ (Python) and js/src/providers/ (JS). If you use a provider we don't support, add a connector for it.
Found a bug? Fix it and submit a PR. If you're not sure how to fix it, open an issue first.
- Python: Follow PEP 8.
- Line length: 100 characters max.
- Type hints: Use type annotations for function signatures.
- Docstrings: Required for public classes and functions.
-
Fork the repository and create a branch from
main:git checkout -b feat/your-feature
-
Make your changes with clear, focused commits.
-
Write tests for new functionality.
-
Push your branch and open a pull request against
main. -
Describe your changes in the PR:
- What does this change?
- How was it tested?
- Any breaking changes?
Use clear, imperative-mood messages:
Add Unicode homoglyph detection for extraction probesFix n-gram scoring for short promptsAdd Google Gemini connector
- Use GitHub Issues for bugs and feature requests.
- Include reproduction steps, expected vs actual behavior.
- For security vulnerabilities, please email hello@agentseal.org instead of opening a public issue.
By contributing, you agree that your contributions will be licensed under the FSL-1.1-Apache-2.0 license.