scientific-llm-benchmarks

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

• Third-party content exposure detected (high risk: 0.85). The required workflow includes fetch_examples.py, which fetches “real sample rows” from HuggingFace dataset URLs at runtime; those dataset rows are outsider-authored free-form text (questions/prompts) that can be inserted into the agent’s LLM context.