Kubernetes Cluster Management

Manage Kubernetes clusters through kubectl and Helm operations. This skill provides comprehensive cluster management capabilities including resource inspection, application deployment, pod troubleshooting, node management, and Helm chart operations.

Prerequisites

Before using this skill, ensure you have:

1. **kubectl** installed and available in your PATH

2. A valid **kubeconfig** file with contexts configured (default: `~/.kube/config`)

3. Access to a Kubernetes cluster (minikube, Rancher Desktop, GKE, EKS, AKS, etc.)

4. **Helm v3** installed (optional, only if using Helm operations)

Verify your connection by running `kubectl get pods` to ensure cluster access works without credential issues.

Core Capabilities

Resource Management

Discovery & Documentation

Advanced Operations

Helm Operations

Troubleshooting

Instructions

1. Basic Resource Inspection

To view cluster resources:

```

Use kubectl_get to list all pods in the default namespace

Use kubectl_get with namespace parameter to list resources in a specific namespace

Use kubectl_describe to get detailed information about a specific pod, deployment, or service

Use kubectl_logs to retrieve logs from a container in a pod

```

2. Application Deployment

To deploy applications to the cluster:

```

Use kubectl_create with a YAML manifest to create new resources

Use kubectl_apply with YAML to create or update resources declaratively

Use kubectl_scale to adjust the number of replicas for a deployment

Use kubectl_rollout to check rollout status or restart a deployment

```

3. Troubleshooting Pods

When pods are failing or misbehaving:

```

Invoke the k8s-diagnose prompt with a keyword (e.g., pod name prefix) and optional namespace

Follow the autonomous troubleshooting flow to identify issues

Use kubectl_logs to examine container logs for errors

Use kubectl_describe to check pod events and status

Use cleanup_pods to remove stuck pods in states like CrashLoopBackOff or ImagePullBackOff

```

4. Helm Chart Management

To deploy and manage Helm charts:

```

Use install_helm_chart to install a chart from a repository

Use upgrade_helm_chart to update an existing release with new values

Use uninstall_helm_chart to remove a Helm release

Use helm_template_apply as an alternative that renders templates and applies with kubectl (avoids auth issues)

Use helm_template_uninstall to uninstall charts using the template approach

```

5. Node Maintenance

To perform node maintenance operations:

```

Use node_management with action "cordon" to mark a node as unschedulable

Use node_management with action "drain" to safely evict pods before maintenance

Use node_management with action "uncordon" to return a node to service

```

6. Debugging and Port Forwarding

To debug applications locally:

```

Use port_forward to forward a local port to a pod or service

Access the application on localhost at the forwarded port

Use stop_port_forward when done to terminate the forwarding session

```

7. Non-Destructive Mode

For read-only or create/update-only access (no deletions):

```

All kubectl_get, kubectl_describe, kubectl_logs operations remain available

kubectl_apply, kubectl_create, kubectl_scale, kubectl_patch, kubectl_rollout remain available

Destructive operations like kubectl_delete, uninstall_helm_chart, and cleanup_pods are disabled

Configure in Claude Desktop by setting env var: ALLOW_ONLY_NON_DESTRUCTIVE_TOOLS=true

```

Security Features

Examples

**List all pods in the kube-system namespace:**

```

Use kubectl_get to list pods in namespace kube-system

```

**Deploy a new application from YAML:**

```

Use kubectl_apply with the following YAML manifest: [provide manifest]

```

**Troubleshoot a failing pod:**

```

Invoke k8s-diagnose prompt with keyword "my-app" and namespace "production"

```

**Scale a deployment to 5 replicas:**

```

Use kubectl_scale to set deployment/my-app to 5 replicas in namespace default

```

**Install a Helm chart:**

```

Use install_helm_chart to install nginx-ingress from the ingress-nginx repository

```

**Forward local port 8080 to a service:**

```

Use port_forward to forward local port 8080 to service/my-service on port 80 in namespace default

```

Constraints