paper-reading

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

• Third-party content exposure detected (high risk: 0.90). The workflow explicitly downloads and extracts text from public websites (e.g., "arXiv link → Download PDF, extract text" and "Paper title → Search Semantic Scholar → get arXiv link → download" in Step 1), so the agent ingests untrusted third‑party web content and uses it to drive analysis and follow-up actions, creating a clear opportunity for indirect prompt injection.