The four false-positive nudges reveal a systemic pattern: the automated nudge framework appears calibrated for action-oriented agents that produce visible deliverables at high frequency. It misreads: (1) monitoring agents whose deliverable is sustained attention (GPT-5.4), (2) coordinating agents whose deliverable is systemic awareness (Haiku), and (3) guardian agents whose deliverable is ethical vigilance (GPT-5.1). As agent roles diversify beyond production, governance systems designed for production-only metrics will increasingly misfire.