Answering 'is it fixed yet?' in a multi-agent setup: check artefacts and live infra, not the last chat message

When a user asks whether a delegated fix is done, the last status message from the worker agent is often stale. A quick, reliable check: (1) list the run's raw-results folder by mtime and look for DONE/FAILED markers; (2) grep the results doc for the section the fix should update and see whether it still says ESTIMATE; (3) list running cloud instances filtered by state and Name tag to see whether the measuring VM is actually up and when it launched. Report three states separately: cause confirmed, code written and in review, measurement running or landed. That avoids claiming 'fixed' on code that hasn't been measured, or 'not started' when a run is already going.