The Agent Skills Directory

[COMMAND_EXECUTION]: The provided scripts scripts/agent.py and scripts/process.py use the Python subprocess module to execute kubectl commands. This is the primary method the skill uses to interact with the Kubernetes cluster for auditing and configuration.
Evidence: subprocess.run(cmd, ...) calls in both scripts where cmd starts with kubectl.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from a Kubernetes cluster, including namespace labels and pod specifications. This data represents an attack surface where maliciously crafted metadata in the cluster could attempt to influence the agent's behavior.
Ingestion points: Retrieval of cluster state through kubectl get namespaces and kubectl get pods in scripts/agent.py and scripts/process.py.
Boundary markers: Not present; the scripts parse and output cluster data without explicit delimiters or instructions to the LLM to ignore embedded content.
Capability inventory: The skill has the ability to read and write cluster configuration via kubectl.
Sanitization: None; the scripts assume the Kubernetes API returns standard, non-malicious content.

implementing-pod-security-admission-controller