msresearch-bioemu

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

• Third-party content exposure detected (high risk: 0.80). Yes — the skill explicitly fetches and ingests public, user-supplied data (UniProt/PDB/AlphaFold and the ColabFold MMseqs2 service) as part of runtime workflows (see server/app.py endpoints such as /api/predict-uniprot and /api/alphafold-structure and the docs/quick-start note about MMseqs2), and that external content is parsed and used to drive analysis and subsequent actions, so it could enable indirect prompt injection.