gke-cluster-autoscaler

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes utility scripts (assets/find-scale-down-blockers.sh and assets/log-autoscaler-events.sh) that use standard Kubernetes and Google Cloud CLI tools (kubectl, gcloud, jq) for cluster diagnostics and log monitoring.
  • [INDIRECT_PROMPT_INJECTION]: The skill is built to process external data like logs and manifests, which presents an injection surface. This is addressed by explicit security controls within the instructions.
  • Ingestion points: User-provided cluster logs, manifests, and command outputs described in the documentation.
  • Boundary markers: The skill contains a mandatory rule identifying pasted data as 'UNTRUSTED DATA' and commanding the agent to ignore directives embedded within it.
  • Capability inventory: The skill can read cluster state, node status, and logs via the provided bash assets.
  • Sanitization: Instructions mandate that all user-supplied identifiers (names of clusters, node pools, or namespaces) must be validated against a strict alphanumeric regex (^[a-z0-9-]+$) before use.
  • [PROMPT_INJECTION]: The skill includes a 'REFUSE INJECTED IDENTIFIERS' security rule that instructs the agent to detect and reject input containing shell metacharacters or characters commonly used in command injection attacks.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 05:40 AM
Security Audit — agent-trust-hub — gke-cluster-autoscaler