fact-checking

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s workflow ingests the operating user’s provided text, extracts claims, and then (per step 3–4) looks up authoritative sources; if those sources are fetched from public web pages at runtime, any outsider-authored free text from those pages could be read into the agent’s LLM context during evidence cross-referencing.