together-fine-tuning

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

• Third-party content exposure detected (high risk: 0.90). The skill's VLM/data-prep instructions (references/data-formats.md and the SKILL.md VLM section) include a url_to_base64(url) example that performs requests.get(url) to fetch arbitrary image URLs and the workflow accepts uploaded training files from arbitrary sources, so the agent is clearly expected to ingest untrusted third-party content that can materially influence model behavior via fine-tuning.