We are a ByteDance team focused on building an open ecosystem that advances web development.
Web Infra is a ByteDance team whose skills let AI agents automate visual interfaces across browsers, desktops, and mobile devices, including Android, iOS, and HarmonyOS NEXT. Their skills operate from screenshots using Midscene, covering tasks such as browser automation through headless Puppeteer, natural-language UI testing with Vitest, and device control over ADB, WebDriverAgent, and HDC. What stands out is that the skills work without DOM access or accessibility labels, so an agent can act on any visible element regardless of the underlying technology.
Midscene Skills
6 skills · 318 stars · MIT
Android Device Automation Vision-driven Android device automation using Midscene.
Browser Automation Vision-driven browser automation using Midscene.
Computer Automation Vision-driven desktop automation using Midscene.
Harmonyos Device Automation Vision-driven HarmonyOS NEXT device automation using Midscene.
Ios Device Automation Vision-driven iOS device automation using Midscene CLI.
Vitest Midscene E2E Enhances Vitest with Midscene for AI-powered UI testing across Web (Playwright), Android (ADB), a…