academic-deep-research

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly mandates automated use of web_search and web_fetch on public sources (e.g., "START: web_search count=20" and "web_fetch on primary sources" in the Phase 3 Tool Usage Sequence and example.md references like fetching TechCrunch and healthcare IT forum pages), meaning the agent will fetch and interpret untrusted public web content as part of its decision-making workflow, which could allow indirect prompt-injection from third-party pages.