tao-train-action-recognition

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill defines workflows for executing NVIDIA TAO Toolkit commands, specifically the action_recognition CLI for training, evaluation, inference, and export tasks.
  • [EXTERNAL_DOWNLOADS]: The documentation includes instructions for downloading and extracting sample dataset archives such as train.tar.gz and test.tar.gz. These are standard requirements for the model's primary training and testing functions.
  • [SAFE]: All identified external resources, including the container image tao_toolkit.pyt and documentation links, are official NVIDIA vendor resources consistent with the skill's purpose.
  • [SAFE]: The skill uses structured YAML templates and JSON schemas to handle model configurations, which is a recommended practice to ensure data integrity and prevent injection during command assembly.
  • [SAFE]: Digital signature metadata (skill.oms.sig) is present to provide integrity verification for the skill's contents, following security best practices.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 05:47 PM
Security Audit — agent-trust-hub — tao-train-action-recognition