agent-browser-cli skill
A skill connects an agent to your Chrome through an extension and CLI for reading pages, managing tabs and verifying actions.
- 649 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The CLI can read cookies and page content, execute JavaScript and submit forms as the signed-in user.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/agent-browser-cliDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The CLI can read cookies and page content, execute JavaScript and submit forms as the signed-in user.
This is third-party code. Review the repository files before installing.
What it does
agent-browser-cli uses an existing Chrome session with its tabs and signed-in sites. An agent can inspect page structure, execute JavaScript and capture screenshots. It also supports navigation, forms and file uploads. A Rust daemon and extension form a local bridge that starts on demand. The skill guides command selection and connection troubleshooting. It does not require a separate test browser, so its access reaches the selected real session.
Who it is for. For developers and QA engineers testing authorized workflows in their own signed-in browser.
Good fit when
- Test a site using an existing login.
- Find the intended tab among open tabs.
- Inspect the DOM and state before submitting a form.
- Compare an action’s result with a page screenshot.
Not a fit when
- You need a sandbox without access to working sites.
- The user has not authorized reading cookies or account contents.
- The browser cannot install the required extension.
Example request
Find my test Chrome tab, read its title and capture it without changing the page.Limitations
Chrome or Chromium needs the bridge extension; installing only the CLI will not connect an agent to tabs. Using WSL requires WSL itself to be version 2.0.0 or later and a connection to Windows Chrome. Local files need a separate extension permission, so disk access is an additional choice. The CLI uses real cookies and executes JavaScript through CDP; begin with a test profile, or the agent can reach working sites and account data.
How to disable. Disable the bridge extension and remove the agent skill. Inspect unfinished forms and uploads before closing working tabs.
FAQ
When does the daemon start?
On demand during a normal command. No running daemon before the first action is not itself a failure.
What does doctor do?
It inspects state without starting the daemon or changing configuration.
Can it open local HTML?
Yes. Extension version 2.1 or later needs file-URL access enabled in Chrome’s extension settings.
How do I inspect a simplified page?
Run `agent-browser-cli scan --text-only` for a text-oriented view of the current page.
Are other browser profiles automatically covered?
Do not assume so. State and permissions apply to the connected profile; actions should target the intended tab.
Discussion
No questions or reviews yet.
Related
SkillsAn agent drives your real logged-in Chrome: a CLI, MCP and an extension with a separate tab group per session
A fast Rust CLI for agent browser automation: accessibility snapshots with element refs, an MCP server and skills
A skill for Amp, Claude Code and Cursor: the agent drives your live Chrome with its open tabs and signed-in accounts through a single debugging toggle
A Rust binary with no MCP server: an agent drives your open Chrome or Edge, and tabs are pinned with target labels.
A CLI and MCP server that lets an agent drive your real Chrome with its logged-in sessions: 103 commands across 36 platforms plus browser automation
BrowserSkill connects agents to signed-in Chrome or Edge and provides reading, clicks and extraction in a separate task window.
README in short
The browser bridge is a separate tool derived from GenericAgent. Several CLI calls share a connection instead of reconnecting to the extension for each command.
README badge
Original SKILL.md
--- name: agent-browser-cli description: 使用 agent-browser-cli 进行浏览器感知与控制、页面交互、截图/PDF、Cookie/CDP 和排障。 --- # agent-browser-cli 使用 `agent-browser-cli` 控制用户真实 Chrome。底层是 Rust daemon + Chrome 扩展桥,保留登录态和 Cookie;不是 Selenium/Playwright。 ## 正常流程直接执行命令 每次开始浏览器任务,不要先做健康检查,直接执行最贴近目标的命令。`tabs` / `open` / `exec` / `scan` 会按需自动启动 daemon;daemon 未常驻是正常状态,不是故障。 ```bash agent-browser-cli open https://example.com agent-browser-cli scan --text-only agent-browser-cli snapshot --limit 200 agent-browser-cli exec --tab <tabId> 'return document.title' ``` 只有目标命令失败、输出明确提示连接异常、扩展未连接、端口不一致、无可用标签页,或用户明确要求排障时,才进入状态检查: ```bash agent-browser-cli status agent-browser-cli doctor agent-browser-cli logs --tail 100 ``` `status` 中 `daemon_not_running` / `running=false` 不能单独作为停止操作的理由;优先继续执行目标命令让 CLI 自动拉起 daemon。`doctor` 只检查状态,不自动启动 daemon、不改配置、不安装 skill。 补充一个容易误判的点:`status` / `doctor` 里看到 `daemon_not_running`,如果此时还没有执行 `tabs` / `open` / `exec` / `scan`,通常只是 daemon 按需常驻而未启动,不代表故障。只有目标命令已经失败,或者日志/输出明确提示端口、扩展、标签页异常时,才进入排障。 ## 打开本地文件 `open` 支持 HTTP(S) URL、绝对 `file://` URL 和本机文件路径。相对路径以执行 CLI 时的当前目录为基准;支持 `~`、中文、空格和符号链接,符号链接会解析到真实文件。所有普通文件类型均可打开,目录会被拒绝。 ```bash agent-browser-cli open ./demo.html agent-browser-cli open ~/Documents/report.pdf agent-browser-cli open 'file:///Users/me/My%20Files/demo.html#intro' agent-browser-cli open ./demo.html --timeout 15 ``` 本地文件要求 Chrome 扩展版本至少为 2.1,并在 `chrome://extensions` → Agent Browser CLI Bridge → 详情中开启“允许访问文件网址”。权限关闭时 `open` 会在创建标签前失败;`status` / `doctor` 会按 Profile 报告 `file_scheme_access`,但该可选能力不影响普通 HTTP(S) 健康状态。