tao-train-fast-foundation-stereo
Pass
Audited by Gen Agent Trust Hub on Jun 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill utilizes official NVIDIA container registries (nvcr.io) and follows established practices for the TAO Toolkit, which is a trusted development environment.\n- [SAFE]: Docker execution instructions include security best practices, such as mapping user IDs via the
--user $(id -u):$(id -g)flag to maintain proper file permissions and avoid running as root inside the container.\n- [COMMAND_EXECUTION]: The skill uses standard CLI tools (docker,depth_net) for its intended purpose of model training and deployment. These operations are transparently documented and do not involve suspicious execution patterns or unauthorized shell access.\n- [DATA_EXFILTRATION]: No unauthorized data access or exfiltration patterns were identified. Data path references are limited to local mounts or cloud storage placeholders (e.g., AWS S3 bucket patterns).\n- [PROMPT_INJECTION]: No evidence of malicious prompt overrides or safety filter bypass attempts was found in the documentation or instructional text.\n- [SAFE]: Analysis of the skill metadata and reference files confirmed that the content is consistent with the stated purpose and does not contain hidden instructions or multi-layer obfuscation.
Audit Metadata