inno-idea-eval

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly runs literature searches against public sources (arxiv, Semantic Scholar, OpenAlex) via search_ai_papers.py (Step 0.5b) and ingests the search results into an Active Novelty Verification Report that is then injected verbatim into persona evidence blocks and used as the primary basis for scoring and pipeline decisions (novelty_grounding_report.txt → Ideation/ideas and logs), so untrusted third‑party content is fetched and directly influences agent behavior.