rag-eval

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow runs scripts/eval/evaluate_rag.py against user-provided dataset files (--dataset-paths → on-disk train.json + corpus/), so outsider-authored free text can be ingested via those files without any “select a specific trusted item” step.