brenda-database

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

• Third-party content exposure detected (high risk: 0.72). The skill calls the BRENDA SOAP API at runtime (e.g., get_km_values / get_reactions in brenda_client), and then parses and propagates BRENDA’s free-text fields like commentary and literature into the agent’s context via returned strings/dicts (e.g., parse_km_entry stores commentary, and workflows print/consume these values). Since BRENDA content is authored by third parties (scientific literature/curators), this is outsider-authored free text entering the LLM context indirectly through API response text.