research-proposal

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

• Third-party content exposure detected (high risk: 0.85). The runtime workflow runs roles/ScholarEval.md, which uses the sibling literature-search tool (lit_search.py) to retrieve and ingest retrieved paper passages/snippets (and may fall back to WebSearch/WebFetch), i.e., outsider-authored free text from public literature into the agent’s LLM context for evidence grounding.