training-llms-megatron

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • Unverifiable Dependencies (SAFE): The skill recommends standard packages such as megatron-core and the official NVIDIA PyTorch Docker image from nvcr.io. These are well-known resources in the machine learning community.
  • Command Execution (SAFE): Commands provided (torchrun, sbatch) are standard for distributed training and are used correctly within the provided script templates.
  • Data Exposure & Exfiltration (SAFE): No sensitive file access or unauthorized network communication patterns were detected. Data paths are appropriately left as placeholders for the user.
  • Indirect Prompt Injection (SAFE): The skill is designed to ingest training data but lacks the high-risk capabilities (like arbitrary file writing or network exfiltration) required to exploit this surface.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:06 PM