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

doca-setup

Pass

Audited by Gen Agent Trust Hub on Jul 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses privileged commands such as sudo tee and sudo mount for configuring system hugepages and the hugetlbfs filesystem, which are required for DOCA runtime performance. These operations are restricted to relevant system paths and are necessary for the skill's primary function.
  • [EXTERNAL_DOWNLOADS]: Workflows include downloading Docker containers from the official NVIDIA Container Registry (nvcr.io/nvidia/doca/doca) and installing standard development dependencies via the OS package manager. These resources are from a trusted vendor and are essential for environment setup.
  • [PROMPT_INJECTION]: The skill implements an indirect prompt injection surface by processing the output of system tools (such as lspci and pkg-config) to drive its internal routing logic. \n
  • Ingestion points: System diagnostic command outputs in TASKS.md (e.g., uname -m, lspci, pkg-config). \n
  • Boundary markers: The agent is instructed to quote observed output back to the user and confirm the routing decision before proceeding. \n
  • Capability inventory: Subprocess execution for system tuning (sudo tee), environment configuration (mount), package installation (apt), and container management (docker). \n
  • Sanitization: Relies on user confirmation and manual observation of command outputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 28, 2026, 05:09 PM
Security Audit — agent-trust-hub — doca-setup