tao-train-segformer

Pass

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is authored by NVIDIA Corporation and exclusively provides configuration templates, JSON schemas, and documentation for the TAO (Train, Adapt, and Optimize) Toolkit.
  • [EXTERNAL_DOWNLOADS]: The skill refers to official NVIDIA Docker images from nvcr.io (NVIDIA GPU Cloud), which is a well-known and trusted container registry.
  • [EXTERNAL_DOWNLOADS]: The skill includes configuration for Weights & Biases (wandb), which is a well-known technology service used for experiment tracking in machine learning.
  • [CREDENTIALS_UNSAFE]: Several spec templates include encryption_key: tlt_encode. This is a publicly documented default placeholder key for the TAO Toolkit and does not constitute a sensitive credential leak.
  • [COMMAND_EXECUTION]: The commands defined in references/skill_info.yaml (e.g., segformer train -e {config_path}) are standard CLI operations for the intended model toolkit.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 23, 2026, 12:09 PM
Security Audit — agent-trust-hub — tao-train-segformer