consume-midi

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

• Third-party content exposure detected (medium risk: 0.30). In search.py, the runtime path /consume-midi → search builds a free-text query from user-supplied filters and sends it to the memory daemon via POST /recall ({"query": query, "k": ...}), so outsider-authored text is ingested by the workflow at recall time.