mcore-testing

Pass

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions for executing tests using torch.distributed.run, pytest, and uv. It also references internal shell scripts like tests/functional_tests/shell_test_utils/run_ci_test.sh for driving functional tests and CI parity.
  • [EXTERNAL_DOWNLOADS]: The skill includes a helper script download_golden_values.py used to fetch reference data from GitHub. Additionally, the unit test configuration in conftest.py is documented to automatically download missing test data from remote sources if not present locally. These operations target well-known repositories associated with the project's development workflow.
  • [SAFE]: All operations described—including network access for data and local script execution—are strictly limited to the intended purpose of facilitating software testing for the Megatron-LM project by NVIDIA.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 3, 2026, 11:04 AM
Security Audit — agent-trust-hub — mcore-testing