The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): The skill documentation includes hardcoded sensitive information in its examples. Evidence: 'postgres://user:pass@host:5432/db' and 'api-key: secret-key-value' in the Secret resource manifest; '--docker-password=pass' in the kubectl command-line example.
[COMMAND_EXECUTION] (HIGH): The skill provides tools for arbitrary command execution within the target environment. Evidence: 'kubectl exec' and 'kubectl debug' allow running interactive shells and arbitrary binaries inside cluster containers.
[PROMPT_INJECTION] (HIGH): There is a high risk of indirect prompt injection (Category 8) due to the combination of untrusted data access and high-privilege write capabilities. * Ingestion points: The agent reads untrusted content via 'kubectl logs' and 'kubectl get events'. * Boundary markers: No markers or instructions are provided to help the agent distinguish data from instructions. * Capability inventory: Includes powerful tools like 'kubectl apply', 'kubectl exec', and 'kubectl edit'. * Sanitization: No sanitization or validation of the ingested cluster data is performed.

kubernetes-ops