AgentDB Vector Search

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s workflow ingests arbitrary user-provided strings as documents or queries passed to AgentDB (e.g., adapter.insertPattern({ ... text }) and db.findSimilar/query/searchSimilar(...)), so outsiders can submit free text that the LLM may embed and use for retrieval without any first selecting a specific trusted item.