The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill facilitates the ingestion of GKE logs and Kubernetes events, which originate from workloads that could be controlled by a third party. This creates a potential surface for indirect prompt injection, where malicious instructions embedded in logs could be misinterpreted as commands by an AI agent. (1) Ingestion points: The skill uses get_k8s_logs, list_k8s_events, and gcloud logging read to import external data. (2) Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands within the log data. (3) Capability inventory: The skill provides access to powerful cluster management tools including gcloud and kubectl. (4) Sanitization: No specific filtering or sanitization of retrieved log content is defined.

gke-observability