knock-in-app-ui

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow instructs the agent to run Knock CLI commands that return guide/message-type JSON (e.g., knock guide list --json, knock message-type get ... --json) and then use that returned content as typed guide data in the LLM context, which is outsider-authored content from the user’s Knock dashboard (not authored by the operating user in this chat).