nielsen-heuristics

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 ingests the user-provided artifact (e.g., a live URL or codebase) and then reads its content into the LLM context; if that artifact includes outsider-authored free text (public web page body, issue/wiki/discussion text, etc.), that text is ingested as readable prose during Step 1/2 and used for Step 3 evidence/locators.