biomedical-skills

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

• Third-party content exposure detected (high risk: 0.80). The package includes multiple skills (e.g., med-researcher / med-researcher-r1, clinicaltrials-api/clinicaltrials-api-v2, ena-sequence-api, genomas, genoTEX, etc.) whose required workflows explicitly query and ingest data from public third‑party sources (PubMed/PMC, ClinicalTrials.gov, ENA/GEO/TCGA, public APIs like AlphaFold/BioThings/Ensembl) and then read/interpret that content to drive hypothesis generation, evidence synthesis, experimental design or downstream actions, which creates a clear vector for indirect prompt injection via untrusted user-submitted/public records.