tao-train-single-step

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

• Third-party content exposure detected (medium risk: 0.65). The required runtime workflow ingests user-provided dataset URIs (e.g., train_dataset_uri / eval_dataset_uri) and then reads the corresponding dataset content for training/evaluation, which is outsider-authored free text/data when the dataset comes from outside the user’s workspace (e.g., public/third-party files fetched via those URIs).