gke-ai-troubleshooting-tpu-vbar-oom

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow reads first-party Cloud Logging entries via query_logs with user-supplied {node_name}/time window to filter serial console and container logs; those log messages are not outsider-authored free text by default, but in practice they could include attacker-controlled payloads if outsiders can write logs/telemetry.