rebuttal

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md’s required workflow explicitly ingests “Raw reviews” (verbatim reviewer text) and then normalizes them into rebuttal/REVIEWS_RAW.md and uses [raw reviews + issue board + draft + venue rules] as part of the Codex MCP stress-test prompt, so outsider-authored free text from reviewers can be read by the agent’s LLM at runtime.