gpu-kubernetes-operations
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPRIVILEGE_ESCALATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches Helm charts and container images from NVIDIA's official registries (
helm.ngc.nvidia.comandnvcr.io). These are established and well-known sources for GPU management software. - [PRIVILEGE_ESCALATION]: Configures a DaemonSet with
privileged: truefor the NVIDIA device plugin. This is a standard and necessary configuration for Kubernetes device plugins to access GPU hardware devices on the host nodes. - [COMMAND_EXECUTION]: Utilizes standard administrative tools including
helm,kubectl, andjqfor cluster configuration, resource labeling, and status verification. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from Kubernetes API queries (
kubectl get ... -o json). This represents a standard operational pattern for cluster management.
Audit Metadata