huggingface-vision-trainer

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/dataset_inspector.py, the runtime workflow fetches and parses outsider-authored dataset rows/fields from the Hugging Face Datasets Server (via the --dataset/--split args) and prints JSON/human-readable previews that can contain arbitrary free text (e.g., JSON-encoded prompt, labels/categories), which the LLM may ingest for decision-making without selecting a specific trusted item.