Get an agent evaluation by ID
Retrieve a single agent evaluation by its UUID
Authorizations
Path Parameters
Response
Agent evaluation details
Latest run, populated via lateral join on eval_evaluation_runs. LatestRunID is what a row-level "stop run" action posts to, and LatestRunCancelRequested lets the row show the pending state during the window between the 202 and the worker writing 'cancelled'. Both are nil when the evaluation has never run.
Latest run carrying a numeric score, via its own lateral. A failed, queued, or running run has no summary, so LatestRunSummary is null for exactly the cases where a score is still worth showing. Consumers read the state from LatestRun* and the numbers from here: folding the two together would hide the failure. Equal to LatestRunID when the newest run is scored, and all nil when no run has ever scored.

