skills/nvidia/skills/run-on-slurm/Gen Agent Trust Hub

run-on-slurm

Pass

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is a technical guide for launching Megatron-LM training jobs on SLURM clusters. All instructions and scripts use standard HPC tooling and do not contain malicious patterns. All resource references (Megatron-LM, Blackwell, Hopper) are consistent with the vendor author context (nvidia).
  • [COMMAND_EXECUTION]: The skill includes bash templates for job submission and task execution within a cluster environment. These commands are necessary for the skill's primary purpose and use placeholders for sensitive or environment-specific data (e.g., <SLURM_ACCOUNT>, <MEGATRON_ARGS>).
  • [EXTERNAL_DOWNLOADS]: Mentions dependency installation via uv sync, which is a standard development practice for Python projects to ensure consistent environments across cluster nodes.
Audit Metadata
Risk Level
SAFE
Analyzed
May 19, 2026, 05:48 PM
Security Audit — agent-trust-hub — run-on-slurm