nemo-mbridge-multi-node-slurm

Pass

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides comprehensive templates and instructions for executing Slurm-related shell commands such as sbatch, srun, and salloc. It also details the use of torch.distributed.run and uv run for launching distributed Python tasks. These commands are essential for the skill's documented purpose of managing high-performance computing jobs.
  • [EXTERNAL_DOWNLOADS]: Guidance is provided for managing Python dependencies within Slurm jobs using pip and uv. The instructions refer to installing packages and syncing environments from the user's defined repositories and shared filesystems. External references target established services like GitHub (specifically NVIDIA's repositories).
  • [CREDENTIALS_UNSAFE]: The skill mentions the use of sensitive tokens (HF_TOKEN, WANDB_API_KEY, GH_TOKEN). Crucially, it includes strong security recommendations against hardcoding these values in scripts and logs, instead advising the use of scheduler environment variables or restricted secrets files.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 12, 2026, 07:42 PM
Security Audit — agent-trust-hub — nemo-mbridge-multi-node-slurm