couchbase-ai-applications

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime RAG workflow (retrieve top-k chunks/memories from Couchbase and then “Pass retrieved chunks + user query to the LLM”) can inject outsider-authored text because retrieved chunk_text / memory content may originate from documents or conversations the operating user did not author.