skills/nvidia/skills/doca-gpi/Gen Agent Trust Hub

doca-gpi

Pass

Audited by Gen Agent Trust Hub on Oct 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided CUDA source code, environment paths, and DOCA error logs to provide programming guidance (SKILL.md, TASKS.md). While it lacks explicit sanitization or boundary markers, it includes no automated execution capabilities. The instructions emphasize verifying preconditions and following the explicit DOCA object model to reduce misconfiguration risks.- [COMMAND_EXECUTION]: The documentation guides the agent to recommend standard infrastructure and diagnostic commands to the user, including pkg-config for build configuration, nvidia-smi for hardware status, and nvcc for CUDA compilation (TASKS.md).- [PRIVILEGE_ESCALATION]: The troubleshooting workflow in TASKS.md mentions the use of sudo dmesg for kernel log analysis. This is a standard procedure for identifying driver-level failures (mlx5/GPU driver) but involves instructing the user to perform privileged operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 16, 2026, 05:17 AM