Screen-Sensing Governance Pattern: An APR-Series Extension for Visual Agent Substrate Governance

The capability for AI agents to capture, interpret, and act on screen-state information introduces a sensory modality with no current governance substrate. Unlike API-mediated agent actions, screen-sensing operates at the pixel layer—below existing access controls, audit logs, and policy enforcement points. This paper defines the Screen-Sensing Governance Pattern (SSGP), an APR-Series extension that places a governed substrate between visual sensors and reasoning agents. The pattern is grounded in a single missing invariant that no current implementation enforces: interpretation is not truth. It is a claim requiring validation, lineage, and authority. The SSGP enforces this invariant through five governance objects, a constrained lifecycle, and a set of non-negotiable substrate invariants across authority, lineage, reversibility, boundary integrity, and drift control.