tao-generate-referring-expressions
Pass
Audited by Gen Agent Trust Hub on Jun 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill facilitates the use of well-known and trusted AI services such as NVIDIA Inference API and Google Gemini for its primary pipeline functions.
- [SAFE]: External setup instructions involve pulling resources from reputable sources, including Docker Hub, PyPI for the vLLM package, and HuggingFace for model weights.
- [SAFE]: API credential handling is implemented using secure practices such as environment variables and placeholders, with no evidence of hardcoded keys or unsafe exposure.
- [SAFE]: Data processing of images and KITTI labels is performed through established TAO Toolkit tools as part of the skill's documented annotation workflow.
Audit Metadata