tao-train-metric-learning-recognition

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill defines several automated actions (train, evaluate, export, inference) that execute the ml_recog CLI tool within a Bash environment.
  • [COMMAND_EXECUTION]: Provides guidance on setting the environment variable TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD=1 to ensure compatibility when loading certain PyTorch checkpoints. The documentation includes a explicit security warning to only apply this setting when handling trusted checkpoints produced by the authorized workflow, preventing unsafe deserialization of untrusted data.
  • [EXTERNAL_DOWNLOADS]: Pulls and runs official NVIDIA TAO Toolkit container images from the NVIDIA Container Registry (nvcr.io). These are verified vendor resources from a well-known service and are used for the primary purpose of the skill.
  • [EXTERNAL_DOWNLOADS]: References dataset ingestion from S3 buckets and remote archives for model training. The skill provides clear per-action dataset requirements and expects standard image classification formats.
  • [SAFE]: The skill implements best practices for secret management by using placeholders and environment variable instructions (e.g., encryption_key) rather than hardcoding credentials.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 05:48 PM
Security Audit — agent-trust-hub — tao-train-metric-learning-recognition