The Agent Skills Directory

Indirect Prompt Injection Surface: The skill ingests data from external sources such as Kubernetes events and Pod status descriptions. This represents a potential consideration where untrusted data from the cluster environment could be used to influence the agent's behavior. However, this is a necessary part of the diagnostic process and is mitigated by the skill's read-only approach and requirement for human review of proposed changes.
Ingestion points: Commands like kubectl describe pod and kubectl get events pull data from the cluster into the agent's context (found in SKILL.md).
Boundary markers: The skill explicitly defines a non-interactive diagnostics boundary.
Capability inventory: The skill uses kubectl and gcloud for environmental discovery and diagnostics.
Sanitization: The skill uses standard output from system tools; no explicit sanitization of these outputs is implemented within the prompt.
Diagnostic Command Execution: The skill executes command-line utilities including gcloud and kubectl to retrieve cluster credentials and inspect resource states. While executing shell commands requires careful management, these are standard diagnostic actions used to gather necessary context for troubleshooting GKE storage issues.
External Resource References: The skill includes links to official documentation on the vendor's official domain (cloud.google.com). These references provide the user with additional technical context and originate from a trusted domain associated with the platform being managed.

gke-storage-troubleshooting