tao-train-dino
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The analyzed skill is a legitimate technical resource for the NVIDIA TAO Toolkit. All documented commands and external references are consistent with the stated purpose of training and deploying machine learning models.
- [COMMAND_EXECUTION]: The skill documents the execution of TAO Toolkit CLI commands (e.g.,
dino train,dino evaluate,dino export) and standard Docker run commands for deployment. These are the primary functions of the skill and are performed within the intended workflow. - [INDIRECT_PROMPT_INJECTION]: The skill defines several ingestion points for untrusted data, specifically dataset URIs (S3 paths) and annotation files (COCO format). While these represent a surface for indirect prompt injection, the risk is mitigated by the structured nature of the data (JSON/YAML) and the specialized execution environment of the TAO Toolkit. This behavior is necessary for the skill's primary purpose.
- [DATA_EXPOSURE]: Technical documentation and templates refer to an
encryption_keywith a value oftlt_encode. This is a well-known default value for the NVIDIA TAO Toolkit and does not constitute a sensitive credential or data exposure risk.
Audit Metadata