patent-risk-checker

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

• Third-party content exposure detected (high risk: 0.85). The skill calls Zhihuiya via the runtime-provided imageUrl and then ingests the returned patent metadata (e.g., title, currentAssignee, legalStatus/event text) into the generated JSON report and chart labels, which are then included in the agent’s LLM context—this is outsider-authored free text from a public patent database API response.