tao-train-mask-grounding-dino

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions and metadata for running the mask_grounding_dino utility to perform machine learning tasks such as training, evaluation, and inference. These operations are restricted to the official NVIDIA toolset.
  • [EXTERNAL_DOWNLOADS]: The skill references official NVIDIA TAO Toolkit container images from the NVIDIA Container Registry (nvcr.io) and standard pre-trained weights from HuggingFace Hub. Both are well-known and trusted technology platforms.
  • [SAFE]: No prompt injection, data exfiltration, or obfuscation techniques were identified. The skill follows established practices for managing ML experiment configurations and relies on secure, vendor-provided infrastructure.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 05:48 PM
Security Audit — agent-trust-hub — tao-train-mask-grounding-dino