Can AI Agents Protect Product Boundaries? Testing Autonomous Workflow Defense

This experiment tackles a critical challenge for AI-driven products: can autonomous agents enforce product boundaries while handling user requests? Using GitHub Issues as a control surface, the author tests both standard workflows and adversarial attacks to determine whether AI can defend product ownership without infrastructure-level controls. The implications are significant—it demonstrates a new pattern for scaling delegated work to AI while maintaining product integrity.