oncall-irm

This skill helps engineers configure alert routing, on-call rotations, and incident response in Grafana IRM and OnCall. It solves the operational problem of getting the right alerts to the right people by covering integrations (Alertmanager, Grafana Alerting, generic webhook, PagerDuty), Jinja2 routing and grouping templates, escalation chains, schedules, Slack ChatOps, and the P1-P4 incident lifecycle. It also flags that OnCall OSS is in maintenance mode (archived March 2026) and that Grafana Cloud users should use IRM, noting the concepts are identical.

The skill is organized around common workflows with verification steps built in: wiring Alertmanager to IRM and previewing the routing template before going live, building an escalation chain (notify on-call, wait, notify team, trigger outgoing webhook, auto-resolve) and testing it with a demo alert, and creating a schedule from an iCal feed then confirming who is on-call via the API. Reference files document Jinja2 routing templates (which return True/False, first match wins), grouping ID templates, advanced template helpers, escalation-chain step types, and iCal/Terraform schedule creation. Best practices include keeping chains shallow with a definitive final step, setting send_resolved so alerts auto-resolve, and assigning integrations and schedules to teams for RBAC.

It is aimed at SRE, platform, and DevOps engineers setting up incident management. The API interactions shown use an Authorization token supplied by the user against the user's own Grafana Cloud stack; the curl calls are standard, mostly read-oriented or resource-creation operations, with credentials handled as normal API tokens rather than being harvested or exfiltrated.