database-lookup

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

• Third-party content exposure detected (high risk: 0.85). At runtime the skill queries multiple public REST APIs (e.g., BindingDB, PubChem, ENA, etc.) and then injects the returned raw JSON (which can include free-text fields like titles/descriptions/synonyms) into the agent’s LLM context; these are outsider-authored dataset content fetched over the network.