benchmark-paper

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md’s required workflow invokes study-exemplars to “fetch … N recent best-paper awardees and top-cited papers … and extract their feature profile” from open-access sources; that runtime-fetched outsider-authored paper text (or extracted prose features/bases) can be ingested into the agent/LLM context for scoring, creating an indirect prompt-injection risk.