Skills
Navigate
skills/learningmatter-mit/atomisticskills/ml-mlip-nvalchemi/Gen Agent Trust Hub

ml-mlip-nvalchemi

Pass

Audited by Gen Agent Trust Hub on Jul 2, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references a machine learning toolkit hosted on an official corporate registry.
  • Evidence: SKILL.md refers to the nvalchemi-toolkit and provides the URL https://pypi.nvidia.com for installation.
  • [DYNAMIC_EXECUTION]: The provided benchmark scripts utilize dynamic importing to load model wrapper classes based on the environment selected by the user.
  • Evidence: In scripts/run_md_benchmark.py and scripts/run_nvalchemi_benchmark.py, the _load_wrapper function uses importlib.import_module to load modules defined in a hardcoded internal registry (ENV_MODELS). This is a legitimate architectural pattern for supporting multiple backend frameworks.
  • [COMMAND_EXECUTION]: The documentation provides standard CLI commands for running internal Python scripts to measure performance.
  • Evidence: SKILL.md Step 6 lists shell commands to execute the local scripts/run_nvalchemi_benchmark.py script with specific flags.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 2, 2026, 02:27 AM