About k8s
k8s is an operational skill for accessing and troubleshooting Kubernetes clusters, aimed at people handling routine cluster maintenance. It provides common kubectl and Flux checks, explains Flux states, and outlines a way to research unfamiliar service documentation.
Key features
- 01Access guidance for dev, integration, and live
- 02kubectl checks for resources, logs, and events
- 03Flux status checks and reconciliation commands
- 04Flux state meanings and suggested actions
- 05Research workflow for unfamiliar services
Use cases
- 01Check pod status, logs, and resource usage
- 02Diagnose issues with Flux-managed deployments
- 03Find failure modes and health indicators for an unfamiliar service
Install
Copy the command and run it in a terminal.
$
npx skills add https:/After installing, type/k8sin Claude Code to confirm it loaded.
Bundled filesservices.mdtests.yaml
Details
From the project READMEACCESSING CLUSTERS
The available clusters are dev, integration, and live. Use the coresponding kube context for access.
Common kubectl Patterns
| Command | Purpose |
|---|---|
kubectl get pods -n <ns> | List pods in a namespace |
kubectl get pods -A | List pods across all namespaces |
kubectl describe pod <pod> -n <ns> | Detailed pod info with events |
kubectl logs <pod> -n <ns> --tail=100 | Recent logs from a pod |
kubectl logs <pod> -n <ns> --previous | Logs from previous container instance |
kubectl get events -n <ns> --sort-by='.lastTimestamp' | Recent events timeline |
kubectl top pods -n <ns> | CPU/memory usage per pod |
kubectl top nodes | CPU/memory usage per node |
kubectl get ns <ns> --show-labels | Namespace labels (network policy profiles) |
kubectl explain <resource> | API schema reference for a resource type |
Flux GitOps Commands
# Check status
flux --context <cluster> get all
flux --context <cluster> get kustomizations
flux --context <cluster> get helmreleases -A
# Trigger reconciliation
flux --context <cluster> reconcile source git flux-system
flux --context <cluster> reconcile kustomization <name>
flux --context <cluster> reconcile helmrelease <name> -n <namespace>
Flux Status Interpretation
| Status | Meaning | Action |
|---|---|---|
Ready: True | Reconciled and healthy | None |
Ready: False | Failed to reconcile | Check the message/reason |
Stalled: True | Stopped retrying after repeated failures | Suspend/resume to reset (see sre skill) |
Suspended: True | Intentionally paused | Resume: flux resume <type> <name> |
Reconciling | Actively being applied | Wait for completion |
Researching Unfamiliar Services
When investigating unknown services, spawn a haiku agent to research documentation:
Task tool:
- subagent_type: "general-purpose"
- model: "haiku"
- prompt: "Research [service] troubleshooting docs. Focus on:
1. Common failure modes
2. Health indicators
3. Configuration gotchas
Start with: [docs-url]"
Chart URL to Docs mapping:
| Chart Source | Documentation |
|---|---|
charts.jetstack.io | cert-manager.io/docs |
charts.longhorn.io | longhorn.io/docs |
grafana.github.io | grafana.com/docs |
prometheus-community.github.io | prometheus.io/docs |
Common Confusions
BAD: Use helm list to check Helm release status
GOOD: Use kubectl get helmrelease -A — Flux manages releases via CRDs, not Helm CLI
Comments
Markdown supported