vk

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

• Third-party content exposure detected (high risk: 0.75). The required workflow calls VK API endpoints at runtime (e.g., groups.getById, wall.post, wall.get) and ingests the returned JSON fields like response/error_msg into the LLM context via jq output; these are outsider-authored free text coming from VK’s public API responses.