Verdicts and evidence

Every goal in a run ends in exactly one verdict, and every verdict is backed by captured evidence: screenshots, accessibility-tree snapshots before and after each action, and the network events observed during the walk.

The verdicts

• completed: the goal was achieved and the evidence proves it. Proof means the specific checkpoint facts, not "the walk finished without an error".
• defect: the walk surfaced behavior that is wrong on its own terms. The finding carries a severity (P1 through P4), the evidence that supports it, and a source pointer when the code locator found a supportable one.
• blocked: the goal could not be completed, and the handoff says exactly what stood in the way. A signup form that rejects every valid input is a blocked verdict naming the dead end. Blocked is a first-class result, not a failure of the tool: discovering that a user cannot get through is precisely the finding.
• inconclusive: the walk finished without a defect, but the run lacked the authority to call it a pass.

Coverage authority

A clean walk only means something relative to what Greenroom was allowed to know about your app.

• With a state contract, coverage authority comes from your declared screen graph: the run can state which changed states were covered, and a clean covered run is a real pass.
• Without one, authority is requested-goal-only: Greenroom walked what it could infer, and a clean run caps at inconclusive, because "we found nothing" is not the same claim as "we covered the change and found nothing".

This cap is deliberate. Greenroom reports what it can prove and labels the boundary, rather than letting a green check imply coverage that never happened.

Fail closed, always

When something breaks mid-run (the app crashes, the environment misbehaves, a model becomes unavailable), the affected goal fails closed as blocked with the cause recorded. A Greenroom handoff never silently omits a goal, and infrastructure trouble is never dressed up as an app defect or a pass.