tao-train-sparse4d

Pass

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill is designed to generate and execute CLI commands for the NVIDIA TAO Toolkit.
  • Execution patterns identified in references/skill_info.yaml include commands like sparse4d train -e {config_path} and sparse4d evaluate -e {config_path}.
  • These commands are run via the Bash tool and are consistent with the skill's stated purpose of training and managing deep learning models.
  • [EXTERNAL_DOWNLOADS]: The skill integrates with external services for data management and telemetry.
  • The documentation in SKILL.md describes the use of S3 buckets (s3://...) for dataset storage and retrieval.
  • The configuration schemas (e.g., schemas/train.schema.json) show that Weights & Biases (WandB) telemetry is enabled by default for experiment tracking, which involves sending training metrics to the WandB cloud service.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 23, 2026, 12:10 PM
Security Audit — agent-trust-hub — tao-train-sparse4d