Adjacent Visibility

Adjacent Visibility is the controlled extension of the agent's perceptual field. The agent does not see only what the current Routine step explicitly names. It can see neighbouring Variables, prior decisions in the trace, related entries in the Glossary component, and policy-permitted memory surfaces. The visibility is authored: the domain expert specifies what counts as adjacent for which decisions. Without authored bounds, adjacency becomes leakage. With authored bounds, adjacency is what lets the agent reason about a customer's full state without violating policy scope. Adjacent Visibility is a primary surface that Dynamic Context Engineering composes against on every decision.