tao-train-ocrnet
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill defines a legitimate workflow for training and deploying optical character recognition (OCR) models using the NVIDIA TAO Toolkit. No malicious patterns such as prompt injection, credential theft, or unauthorized persistence were found.
- [EXTERNAL_DOWNLOADS]: The skill references official NVIDIA TAO Toolkit container images (e.g.,
nvcr.io/nvidia/tao/tao-toolkit). These are hosted on the NVIDIA GPU Cloud (NGC) registry, which is a well-known and trusted vendor source for AI development tools. - [COMMAND_EXECUTION]: The skill instructs the agent to execute standard model-specific CLI tools (e.g.,
ocrnet train,ocrnet evaluate) inside Docker containers. This command execution is essential for the skill's intended purpose of model lifecycle management. - [INDIRECT_PROMPT_INJECTION]: The skill processes external datasets consisting of images and ground truth text files. While this represents a potential attack surface for indirect prompt injection, the risk is minimal given the specialized nature of the OCR training process and the isolation provided by containerized execution.
Audit Metadata