EKS agent
Cluster and node health, add-on lifecycle, upgrades, pod scheduling, CrashLoopBackOff, ImagePullBackOff, eviction, and resource pressure.
Diagnostic approach
For node failures inspect conditions, kubelet, network, and disk. For Pending pods inspect scheduler events, selectors, taints, capacity, and volumes. For crash loops inspect previous logs, limits, and configuration. Before upgrades check the target version, add-ons, and deprecated APIs.
Initial read-only checks
Run only in the intended account, region, and Kubernetes context. Service-specific follow-ups come from the observed result.
kubectl cluster-info
kubectl get nodes -o wide
kubectl get pods -A -o wide
kubectl get events -A --sort-by=.lastTimestamp
Evidence and handoff
Return the affected component, observed symptoms, supporting output, likely cause, proposed action, and a verification command with expected results. In team mode, report only the assigned domain and identify dependencies for the coordinator.
The plugin bundles awsdocs and awsapi; additional knowledge/IaC integrations are optional host capabilities. Models and tools are defined in the actual agent frontmatter and Codex overlay, not by a second public-site settings table.