tao-train-ocdnet
Pass
Audited by Gen Agent Trust Hub on Jun 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a configuration-heavy extension for the NVIDIA TAO Toolkit. It defines interfaces (via JSON schemas and YAML templates) for managing machine learning workflows.
- [DATA_EXPOSURE]: No sensitive information or hardcoded credentials were found. Example configurations use placeholder AWS S3 paths (e.g.,
s3://bucket/data/train) as expected for model training guidance. - [EXTERNAL_DOWNLOADS]: The skill references container images from the NVIDIA Container Registry (
nvcr.io), which is an official and trusted service for these tools. - [COMMAND_EXECUTION]: The commands described in the reference documentation and skill metadata are standard TAO CLI operations (e.g.,
ocdnet train,ocdnet evaluate) intended to run within isolated Docker containers. - [INDIRECT_PROMPT_INJECTION]: While the skill defines a surface for ingesting user data paths and specification overrides, it follows standard patterns for model training skills and does not exhibit unsafe interpolation or execution patterns that would facilitate injection attacks.
Audit Metadata