article-to-habr

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

• Third-party content exposure detected (medium risk: 0.30). In “Review” mode the workflow explicitly “read[s] the user’s text” (which is outsider-authored free text) and then runs it through local checklist/guidance sections to produce the review, so the LLM ingests user-provided arbitrary text at runtime.