scientific-writing

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime workflow explicitly depends on the separate research-lookup skill to gather literature and extract key findings for Stage 1 outlines; if that lookup fetches public web content or ingests outsider-authored paper text, those free-form excerpts can be fed into the agent’s LLM context via the outline-to-paragraph writing path.