a2a-redteam

# a2a-redteam A2A Redteam: A simulator for adversarial testing of multi-agent AI systems, exploring rogue agent behavior, unsafe context propagation, and deception resilience. ## Agent Prototype This project implements a minimal agent prototype using FastAPI that: 1. Exposes a `/skills` endpoint that returns the agent's capabilities 2. Exposes a `/execute` endpoint that runs modules with specified targets 3. Logs all activities and returns execution results ### Supported Skills - `scan_nmap`: Network scanning using nmap - `fuzz_ffuf`: Web fuzzing (currently simulated) ### API Endpoints - `GET /`: Returns a welcome message - `GET /skills`: Returns the agent's ID and available skills - `POST /execute`: Executes a module with a target - `GET /history`: Returns execution history ### Setup and Run 1. Install dependencies: ``` python3 -m venv venv source venv/bin/activate pip install -r requirements.txt ``` 2. Run the server: ``` uvicorn main:app --reload ``` 3. Access the API at http://127.0.0.1:8000/ 4. API documentation available at http://127.0.0.1:8000/docs ### Example Usage ```bash # Get agent skills curl -X GET http://127.0.0.1:8000/skills # Execute a module curl -X POST http://127.0.0.1:8000/execute \ -H "Content-Type: application/json" \ -d '{"module": "scan_nmap", "target": "example.com"}' # View execution history curl -X GET http://127.0.0.1:8000/history ```