The AI Security Agent is an automated red-teaming framework designed to test and evaluate the robustness of AI systems against adversarial attacks. It employs a multi-agent architecture where specialized agents collaborate to:

1. Generate adversarial prompts targeting specific vulnerability categories
2. Execute attacks against a target system (e.g., a banking assistant)
3. Evaluate responses to determine if safety guidelines were violated