gke-ai-troubleshooting-tpu-performance-degradation
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection Surface: The skill involves the agent processing outputs from commands such as
gcloud alpha mldiagnostics monitored-events describeandkubectl get nodes. This data ingestion represents a potential surface where the agent interprets external system information to determine the next steps in a troubleshooting path. - Diagnostic Command Generation: The agent is instructed to generate commands for the Google Cloud SDK (
gcloud) and Kubernetes CLI (kubectl). These tools are used within their intended scope for infrastructure monitoring and telemetry analysis, as described in the provided technical documentation. - Official API Utilization: The workflow relies on communication with official Google Cloud service endpoints, including
hypercomputecluster.googleapis.comandmonitoring.googleapis.com, to retrieve workload monitoring events and system metrics. These are standard resources for GKE management and monitoring.
Audit Metadata