tao-run-automl-deft-pipeline
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill manages a canonical NVIDIA AOI training workflow, delegating tasks to established AutoML and DEFT skills. The logic is consistent with standard ML development practices.
- [SAFE]: External dependencies and container images are sourced from trusted NVIDIA domains, including the NVIDIA Container Registry (nvcr.io) and the official TAO Skill Bank on GitHub.
- [SAFE]: The skill implements a consolidated pre-flight process that checks for the presence of required environment variables without logging or displaying their sensitive values.
- [SAFE]: Data processing involves standard ML datasets (CSVs and checkpoints), and the skill provides quality checks to prevent common failure modes like mode collapse on imbalanced data.
- [SAFE]: All command execution and file operations are performed within the expected workspace and are restricted to the primary purpose of model training and optimization.
Audit Metadata