The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains multiple examples of kubectl commands used to manage cluster resources, including apply, delete, scale, and rollout operations.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon Kubernetes manifests and configuration changes, which creates a vulnerability surface if the agent is instructed to process data from untrusted external sources.
Ingestion points: Processes Kubernetes resource definitions (YAML) and configuration updates (images, environment variables) via kubectl apply, kubectl patch, and kubectl set.
Boundary markers: The instructions do not provide specific delimiters or warnings to ignore potentially malicious instructions embedded within user-provided manifests or documentation.
Capability inventory: Extensive cluster management capabilities including resource creation, deletion, and modification via the kubectl CLI.
Sanitization: No validation or sanitization logic is suggested for the Kubernetes manifests or container image names before they are applied to the cluster.

deployment