databricks-apps-python

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

• Third-party content exposure detected (high risk: 0.85). Outsider free text can enter the LLM context via Streamlit st.chat_input() (user-authored prompt) and is appended to st.session_state.messages, which is then sent to client.chat.completions.create(...) as messages=st.session_state.messages in the foundation-model chat examples.