The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the kubectl binary to perform cluster operations. This includes the ability to run arbitrary commands within containers via kubectl exec and to apply configuration changes that can modify the state of the entire cluster.
[DATA_EXFILTRATION]: The skill provides commands that access sensitive configuration files and cluster data. Specifically, it facilitates viewing the kubeconfig file (typically located at ~/.kube/config) and retrieving Kubernetes secrets, both of which contain authentication tokens and credentials required for cluster access.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it retrieves and processes untrusted data from the cluster without proper isolation or sanitization.
Ingestion points: Untrusted data enters the agent context through kubectl logs, kubectl get events, and kubectl describe commands, which fetch content (such as log messages or event reasons) from the cluster environment.
Boundary markers: No delimiters or instructions are used to distinguish cluster-provided data from system instructions.
Capability inventory: The skill possesses broad capabilities, including the ability to execute code in containers, manipulate cluster resources, and access the local file system via kubectl cp.
Sanitization: There is no evidence of sanitization or filtering of the data retrieved from the cluster before it is presented to the agent.

kubectl-skill