together-fine-tuning

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

• Third-party content exposure detected (high risk: 0.90). The vlm_finetune.py workflow (url_to_base64 and the --sample-image-url option) explicitly downloads arbitrary image URLs via requests.get and embeds the resulting data URIs into training and test datasets that the agent will ingest for fine-tuning/inference, so untrusted third‑party content can influence model behavior.