paper-research-assistant

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

• Third-party content exposure detected (high risk: 0.75). Runtime LLM context is fed from parse_paper.py extracting outsider-authored PDF text (user-supplied paper from arXiv/journal or arbitrary PDF path) into full_text_preview/metadata JSON, which generate_report.py then inserts verbatim (e.g., metadata['abstract']) into the generated report prompt/context.