rag-engineer

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

• Third-party content exposure detected (high risk: 0.75). The skill is a RAG pipeline design that ingests external documents (e.g., via runtime document parsing from URLs/files) and then assembles retrieved chunk text into the LLM prompt context under <context>...</context>, so outsider-authored free text can be ingested and fed to the agent’s LLM at runtime.