The Agent Skills Directory

[COMMAND_EXECUTION]: Enables administrative tasks such as executing commands inside containers using kubectl exec and managing cluster workloads.\n- [EXTERNAL_DOWNLOADS]: Fetches official configuration, binaries, and repository keys from Kubernetes domains (dl.k8s.io, pkgs.k8s.io) during the setup process.\n- [EXTERNAL_DOWNLOADS]: Documents the standard kubectl feature for applying manifests directly from remote URLs.\n- [CREDENTIALS_UNSAFE]: Provides examples for viewing and decoding Kubernetes secrets, which is a necessary function for cluster administration and debugging.\n- [PROMPT_INJECTION]: Faces potential exposure to indirect prompt injection by reading data from the cluster (e.g., logs and events) that could be influenced by external actors.\n
Ingestion points: kubectl logs, kubectl describe, and kubectl get events pull data into the agent's context.\n
Boundary markers: The instructions do not define specific markers to isolate the output of cluster-read commands from the agent's instructional logic.\n
Capability inventory: The skill possesses extensive capabilities for cluster resource modification, deletion, and shell access.\n
Sanitization: No explicit sanitization of cluster-provided text (like pod logs) is described in the workflows.

kubectl