@web-infra-dev

chrome-bridge-automation uses screenshots to drive browser interactions via chrome devtools protocol and midscene extension, preserving session state without dom access or direct input device control.

desktop-computer-automation orchestrates vision-driven ui interactions via midscene, enforcing synchronous execution and screenshot-driven decision loops to control local or remote windows desktops through natural language commands.

vision-driven browser automation using headless puppeteer with natural language prompts, no dom access or selectors required. requires vision-capable model configuration and synchronous command sequencing.

harmonyos-device-automation wraps midscene cli commands for screen-driven automation on harmonyos next devices, with strict synchronous execution rules and image-based action selection, but lacks clear intent mapping and error recovery.

ios-device-automation wraps the Midscene CLI to automate iOS device interactions through screenshot analysis. no dom or accessibility labels required. covers basic visual navigation, tapping, typing, and scrolling primitives.

Vision-driven Android device automation using Midscene. Operates entirely from screenshots — no DOM or accessibility labels required. Can interact with all…

vitest-midscene-e2e — an installable skill for AI agents, published by web-infra-dev/midscene-skills.