target-prioritization

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

• Third-party content exposure detected (high risk: 0.80). This skill programmatically fetches and ingests public third‑party content from UniProt, OpenTargets, and PubMed (see scripts/fetch_uniprot.py, scripts/fetch_opentargets.py, scripts/fetch_pubmed.py and SKILL.md), and the LLM (Claude) is explicitly instructed to read those fetched dossier fields to write rationales and drive prioritization/next‑step recommendations, so external content could indirectly influence agent behavior.