evidence-case-viewer

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

• Third-party content exposure detected (medium risk: 0.30). In bridge.py, the GUI chat well takes user-provided free text from QML (sendChat Slot → message) and sends it into the agent prompt/context and then streams the subagent’s response from POST {subagent_url}/chat/stream, so outsider-authored text is ingested at runtime.