What problem does it solve? When a Materialize environment is unhealthy, slow, lagging, restarting, or refusing connections, operators need a disciplined way to investigate using the monitoring stack rather than a SQL session. This Skill routes you to the right troubleshooting documentation and enforces diagnostic habits that keep an investigation honest. ## Core Features & Use Cases - Symptom-based routing: Distinguishes between a broken monitoring stack and a broken Materialize cluster, directing you to the correct troubleshooting guide for each. - Query-first diagnosis: Teaches you to confirm labels with count by (<label>) (<metric>), consult the query registry in packages/queries/*.yaml for written metric interpretations, and return findings as re-runnable expressions instead of screenshots. - No-privilege investigation: Uses gcx to reach Grafana, Thanos, and Loki without Kubernetes cluster-admin access, so diagnosis never requires escalating privileges. - Use Case: Grafana shows a cluster pinned at its memory limit. Use this Skill to identify which dashboard answers the question, verify the scoping label materialize_cloud_organization_name, and produce a query with window and scope that a teammate can re-run. ## Quick Start Ask the assistant to troubleshoot why your Materialize cluster is lagging using the monitoring stack dashboards and queries.
