← ClaudeAtlas

agent-browserlisted

Drive a real browser from the CLI — navigate, snapshot, click, fill, screenshot, emulate. Use when the user wants to interact with a website, test or review a web UI, take screenshots, extract page data, or attach to a CDP endpoint (Chrome, Tauri/WebView2, Electron), or when another skill needs browser evidence.
Firzus/agent-skills · ★ 1 · Web & Frontend · score 72
Install: claude install-skill Firzus/agent-skills
# agent-browser Browser automation CLI built for agents ([vercel-labs/agent-browser](https://github.com/vercel-labs/agent-browser), Apache-2.0): a persistent daemon drives Chrome over CDP, and accessibility-tree snapshots give compact `@eN` refs to act on — no Playwright, no HTML parsing. ```bash npm i -g agent-browser && agent-browser install # once; downloads Chrome for Testing agent-browser doctor # when anything misbehaves ``` ## The core loop ```bash agent-browser open <url> # 1. open a page agent-browser snapshot -i # 2. see it — interactive elements with @eN refs agent-browser click @e3 # 3. act on a ref (click, fill, type, select…) agent-browser snapshot -i # 4. re-snapshot after ANY page change ``` Refs go **stale the moment the page changes** — navigation, submit, re-render, dialog. Re-snapshot before the next ref interaction. When refs fail, fall back to semantic locators (`find role button click --name "Submit"`, `find label "Email" fill "…"`), then raw CSS selectors. The browser persists across commands; `agent-browser close` when done. ## Waiting Bad waits cause more failures than bad selectors. After any page-changing action, pick one: ```bash agent-browser wait @e1 # element appears agent-browser wait --text "Success" # text appears agent-browser wait --url "**/dash" # URL matches (glob) agent-browser wait --load networkidle # SPA catch-all agent-browser wait --fn "windo