A self-contained adversarial simulation platform where AI agents attack and defend each other to test robustness, security, alignment, and reasoning integrity.