paper-audit

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly ingests external literature-search results (e.g., arXiv / S2 / "Tavily") as part of Phase 0 when --literature-search is enabled — see Mode Guide and the literature_reviewer_agent/CHANGELOG references — and those third-party results are read and used by reviewer agents to influence findings and gate decisions, so untrusted public content can materially affect the agent's actions.