The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches installation manifests and management tools from the official GitHub repositories and domains of well-known infrastructure projects, including ArgoCD, FluxCD, Istio, Linkerd, and the Kubernetes Cluster API.
[REMOTE_CODE_EXECUTION]: Instructions are provided for downloading and executing installation scripts for service meshes and cluster networking tools (e.g., Istio's download script and Linkerd's installation script) through piped shell commands.
[CREDENTIALS_UNSAFE]: The documentation contains template examples for Kubernetes Secret resources. These examples use placeholder strings for sensitive information such as OpenAI API keys (sk-...), SSH private keys, and TLS certificates to illustrate secure configuration practices.
[COMMAND_EXECUTION]: The skill facilitates the execution of administrative commands that modify cluster state, including pod debugging on nodes, scaling deployments, and applying complex Helm charts.
[DATA_EXFILTRATION]: Provides standard commands for retrieving and decoding sensitive cluster data, such as using kubectl to output secrets in base64 or streaming logs from production containers for troubleshooting purposes.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the analysis of untrusted data such as container logs or third-party Helm values.
Ingestion points: kubectl logs, kubectl get, and external Git repositories (references/troubleshooting.md, references/gitops.md).
Boundary markers: Absent; no instructions are provided to the agent to isolate or ignore instructions potentially embedded in logs or manifests.
Capability inventory: kubectl apply, kubectl exec, kubectl debug, and helm install across multiple scripts.
Sanitization: Absent; the skill relies on the standard behavior of CLI tools without additional filtering of content derived from external sources.

kubernetes-specialist