paper-workbench

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

• Third-party content exposure detected (medium risk: 0.30). At runtime, when the user supplies a paper landing page / remote PDF URL / arXiv/AlphaXiv/DOI (outsider-authored content), scripts/normalize_paper.py fetches and parses the resulting HTML/PDF/text (including content.full_text_markdown and content.page_chunks.text) via http_get(...), extract_text_from_downloaded_pdf(...), and normalize_arxiv_source(...), so the agent ingests outsider-authored free text.