Did the agent do the right thing, and how do we know?
One task_id is stamped on the intent, the tool calls, the commits, the pipeline runs and the deploy. The last event is not a green build: it is the service confirming the change did what it was meant to.
Requirement
Observability has to answer a new question. Not just "is the service healthy" but "did the agent do the right thing, and how do we know."
Design choices
task_id propagates through commits, PR labels, gateway audit rows, Argo annotations and traces.
Outcome metrics: first-attempt gate pass rate, retries per task, human override rate, rollback rate against a human baseline.
The closing event is a service-level check, not a pipeline status.
Evidence and gap
Built in Hangar: release-outcome-span tracing, the Tower Release Record with compare view, and a HolmesGPT triage prototype. Proposal only: task_id as a first-class field across the gateway and Argo.