tao-train-depth-anything-v2
Pass
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security issues were detected. Analysis confirmed the skill's operations align with its stated purpose of model training and deployment.
- [EXTERNAL_DOWNLOADS]: The skill references pretrained model weight files hosted on Meta's official file distribution domain (dl.fbaipublicfiles.com). This is a standard and safe procedure for initializing neural network backbones in computer vision tasks.
- [COMMAND_EXECUTION]: Command patterns identified in the skill are restricted to the TAO Toolkit CLI (depth_net) and Docker management. These are used strictly for training, inference, and model conversion as part of the intended monocular depth estimation workflow.
Audit Metadata