nemo-mbridge-recipe-recommender

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous shell command templates for running distributed training using torch.distributed.run and benchmarking scripts. These commands reference local project files such as scripts/training/run_recipe.py and scripts/performance/run_script.py.
  • [EXTERNAL_DOWNLOADS]: The skill contains references to official documentation and source code hosted on GitHub (NVIDIA-NeMo/Megatron-Bridge). These links point to well-known technology service domains.
  • [PROMPT_INJECTION]: The instructions guide the agent to collect user parameters (model name, GPU count, training goal) to recommend appropriate recipes. No patterns attempting to bypass safety filters or override system constraints were detected.
  • [SAFE]: The skill is a documentation-centric utility designed to help users navigate complex training configurations. All provided commands and links are consistent with the stated purpose of assisting with Megatron Bridge setups.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 05:11 AM
Security Audit — agent-trust-hub — nemo-mbridge-recipe-recommender