research-recipes

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

• Third-party content exposure detected (high risk: 1.00). The skill autonomously fetches and ingests content from open/public third-party sources (e.g., PubMed, OpenAlex, Semantic Scholar, cBioPortal, ClinicalTrials.gov, AlphaFold, DrugBank via web_search) as required steps in SKILL.md/REFERENCE.md and uses those results to drive analyses and decisions (reports, rankings, follow-up actions), so untrusted third-party content can materially influence tool use and next actions.