Keywords: defensive-publication, prior-art, ai-agents, adaptive-autonomy, human-in-the-loop, confidence-calibration, agent-safety