autodesign

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

• Third-party content exposure detected (high risk: 0.72). Yes: during required runtime iteration steps, the agent calls Figma MCP get_screenshot / get_design_context and then feeds the resulting readable design context/screenshot-derived text into the LLM, where that content originates from the user-selected Figma file (an external/outsider-authored source within the listed category “Documents … authored by outsiders”; plus the user-provided --target is also free text).