feynman-teachback

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

• Third-party content exposure detected (high risk: 0.85). The required workflow has the learner (an outsider relative to the mentor/operating user) produce free-form “learner’s words” and a “verbatim” analogy that are then written into the teach-back Markdown and used as LLM context during the session (listen/probe + capture/condense + record verbatim), which is an indirect prompt-injection risk from outsider-authored text.