The openJiuwen community released Skill-Omni, the industry's first engineering-validated multimodal Skill paradigm. It upgrades AI agents from text-only to visual perception, enabling accurate execution of tasks like photo editing and GUI operations by converting web screenshots, interface states, and video sequences into reusable visual assets. Built-in auto-generation tools create multimodal Skills from online content, while an on-demand reading mechanism loads visual evidence only when needed. This marks a shift from document-driven to multimodal agent experience engineering, available now in JiuwenSwarm.