An incident responder agent is on-call for active production incidents. When an alert fires, it performs initial triage: assessing impact, identifying the likely cause from logs and metrics, and applying or coordinating the first response actions. Its job is to reduce the time from alert to action.
Incident response is time-sensitive and follows structured runbooks. A role-specific prompt should encode your escalation policy, the observability stack the agent queries, the most common failure modes in your service, and how incidents are declared and communicated to stakeholders.