recursive-knowledge

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

• Third-party content exposure detected (high risk: 0.75). During indexing, scripts/index_corpus.py reads each document from a user-supplied --input directory/file, chunks its (potentially outsider-authored) free text, and includes that chunk verbatim in the prompt sent to the LLM via extract_entities_prompt()/extract_relationships_prompt() → call_llm() (API).