tao-train-fast-foundation-stereo

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). At runtime, the skill’s required workflow ingests user-provided dataset annotation text files via dataset.*.data_sources[*].data_file (e.g., dataset.test_dataset.data_sources[0].data_file for evaluate and dataset.infer_dataset.data_sources[0].data_file for inference), which are outsider-authored free-form text unless the operating user authored them.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill's runtime examples and deploy workflow instruct pulling and running the external Docker image nvcr.io/nvidia/tao/tao-toolkit:6.26.3-deploy (nvcr.io/nvidia/tao/tao-toolkit:6.26.3-deploy), which is fetched at runtime and will execute remote code as the required container image for deploy actions.