restaurant-respond

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime workflow explicitly ingests “verbatim review text” from outsider-authored restaurant reviews (e.g., via WebSearch or reading RESTAURANT-REVIEWS-[Name].md), which are then fed into the agent/LLM context for draft generation—this is indirect prompt injection risk from public/authored review text.