gke-ai-troubleshooting-tpu-reservation-stockout
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFE
Full Analysis
- Read-Only Diagnostic Posture: The skill includes an explicit 'Read-only rule' that restricts the agent to diagnostic commands. Any high-risk actions, such as deleting or updating node pools, are designated as user-driven tasks, which maintains a safe human-in-the-loop workflow.
- Vendor-Consistent Tooling: The instructions utilize standard Google Cloud tools (gcloud, kubectl) and metrics consistent with the vendor's (google) official ecosystem. All external references point to legitimate documentation domains (docs.cloud.google.com).
- Prerequisite Role Transparency: The skill clearly outlines the specific IAM roles required for the investigation, promoting the principle of least privilege by ensuring users are aware of the access levels needed before execution.
- Standard Diagnostic Procedures: The workflow follows established troubleshooting patterns for Kubernetes and Compute Engine resources, such as querying audit logs and checking resource utilization via official APIs.
Audit Metadata