mini-browser
Browser automation skill for AI agents using the mb CLI. Use when the agent needs to browse the web, take screenshots, scrape text, fill forms, click elements, record screencasts, run JS in pages, or audit designs. Triggers on: "browse", "open a page", "take a screenshot", "scrape", "fill form", "click button", "web automation", "record screen", "design audit", "accessibility check".
- Installs
- 2.4K
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/runablehq/mini-browser --skill mini-browser- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads mini-browser automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
mini-browser (mb) — Browser CLI for Agents
mb is a browser CLI where each command is a small Unix tool. It talks to Chrome over CDP (port 9222) via puppeteer-core.
Setup (only if not already available)
Setup is only needed when mb is not installed or Chrome is not reachable.
Run these checks first — if both pass, skip straight to the Command Reference.
Check if ready
# 1. Is mb installed?
which mb && echo "mb: ok" || echo "mb: MISSING"
# 2. Is Chrome listening on CDP?
curl -sf http://127.0.0.1:9222/json/version > /dev/null && echo "chrome: ok" || echo "chrome: NOT RUNNING"
If both print "ok", everything is ready — go use mb commands directly.
Install (only if mb is missing)
npm install -g @runablehq/mini-browser
Start Chrome (only if not running)
mb-start-chrome
This launches Chrome with --remote-debugging-port=9222, a fresh profile, and a
1024×768 window. It no-ops if Chrome is already running.
To kill and relaunch:
mb-restart-chrome
Verify
mb go "https://example.com" && mb text
Environment Variables
| Variable | Default | Description |
|---|---|---|
CHROME_PORT | 9222 | CDP port |
CHROME_BIN | auto-detected | Path to Chrome/Chromium binary |
CHROME_PID_FILE | <scripts>/.chrome-pid | PID file location |
CHROME_USER_DATA_DIR | <scripts>/.chrome-profile | Chrome profile directory |
Command Reference
Navigation
| Command | Description |
|---|---|
mb go <url> | Navigate to URL (waits for networkidle) |
mb url | Print current URL |
mb back | Go back |
mb forward | Go forward |
Observation
| Command | Description |
|---|---|
mb text [selector] | Visible text content (default: body) |
mb shot [file] | Screenshot to PNG (default: ./shot.png) |
mb snap | List interactive elements with coordinates |
Interaction
| Command | Description |
|---|---|
mb click <x> <y> | Click at coordinates |
mb type [x y] <text> | Type text (with coords: selects first) |
mb fill <k=v...> | Fill form fields by label/name/placeholder |
mb key <key...> | Press keys (Enter, Tab, Meta+a) |
mb move <x> <y> | Hover at coordinates |
mb drag <x1> <y1> <x2> <y2> | Drag between points |
mb scroll [dir] [px] | Scroll (default: down 500) |
Recording
| Command | Description |
|---|---|
mb record start <file> | Start recording (.webm, .mp4, .gif) |
mb record stop | Stop recording and save |
mb record status | Check if recording is active |
Tabs
| Command | Description |
|---|---|
mb tab list | List open tabs |
mb tab new [url] | Open new tab, print index |
mb tab close [n] | Close tab (default: last) |
Other
| Command | Description |
|---|---|
mb js <code> | Run JavaScript in page context |
mb wait <target> | Wait for ms / selector / networkidle / url:pattern |
mb audit | Design audit (palette, typography, contrast, a11y, SEO) |
mb logs | Stream console logs (Ctrl+C to stop) |
Flags
| Flag | Default | Description |
|---|---|---|
--timeout <ms> | 30000 | Command timeout |
--tab <n> | 0 | Target tab index |
--json | false | Structured JSON output |
--right | false | Right-click |
--double | false | Double-click |
--fps <n> | 30 | Recording frame rate |
--scale <n> | 1 | Recording scale factor |
Usage Patterns
Observe → Act loop
The standard agent loop: snapshot the page, pick an element, act on it.
mb snap # list interactive elements with (x, y)
mb click 512 380 # click the button at those coordinates
mb wait networkidle # wait for the page to settle
mb snap # observe again
Fill and submit a form
mb go "https://example.com/login"
mb fill "Email=user@example.com" "Password=hunter2"
mb key Enter
mb wait url:/dashboard
Take a screenshot
mb shot page.png
mb shot page.png --width 1440 --height 900
Extract text
mb text "main" # text from <main>
mb text "#content" # text from #content
mb text # full body text
Run JavaScript
mb js 'document.title'
echo 'document.querySelectorAll("a").length' | mb js -
Record a screencast
mb record start demo.mp4 --fps 30 --scale 1
# ... interact with the page ...
mb record stop
Design audit
mb audit # human-readable report
mb audit --json # structured JSON output
Dismiss overlays
Cookie banners and modals block clicks. Remove them with JS:
mb js 'document.querySelector("[class*=cookie]")?.remove()'
Wait strategies
mb wait 2000 # sleep 2 seconds
mb wait ".modal" # wait for selector to appear
mb wait networkidle # wait for no network activity
mb wait url:/dashboard # wait for URL to contain string
Important Notes
- Viewport is 1024×768.
snaponly returns elements in the current viewport — scroll and snap again to find more. textuses querySelector — returns first match only. Usetext "main"overtext "p"for better results.gowaits for networkidle. For heavy SPAs, follow up withwait ".selector".typewith coordinates triple-clicks first to select existing text, then types the replacement.fillfield matching order: aria-label → placeholder → name attr → id → label text → CSS selector (use#/./[prefix).--jsonoutput:snap→[{role, name, x, y, state}],tab list→[{index, url, title}],logs→ JSON lines,audit→ full audit object.- Recording state is stored in
~/.mb-recorder.json. Only one recording at a time. tab closecannot close the last remaining tab.
Troubleshooting
| Problem | Fix |
|---|---|
| "Chrome not found" | Set CHROME_BIN=/path/to/chrome |
| Connection refused | Run mb-start-chrome first |
| Stale recording state | Delete ~/.mb-recorder.json |
| Chrome window wrong size | mb-restart-chrome (creates fresh profile) |
| Element not in snap output | mb scroll down 500 then mb snap again |
FAQ
What does the mini-browser skill do?
Browser automation skill for AI agents using the mb CLI. Use when the agent needs to browse the web, take screenshots, scrape text, fill forms, click elements, record screencasts, run JS in pages, or audit designs. Triggers on: "browse", "open a page", "take a screenshot", "scrape", "fill form", "click button", "web automation", "record screen", "design audit", "accessibility check".
How do I install mini-browser?
Run npx skills add https://github.com/runablehq/mini-browser --skill mini-browser in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/mini-browser/ (personal) or .claude/skills/mini-browser/ (project).
Which AI agents can use mini-browser?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is mini-browser safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains mini-browser?
It is published by runablehq in runablehq/mini-browser. It has 2,436 recorded installs.
Popular productivity & automation skills
agent-browser
vercel-labs/agent-browser
Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. Also use for exploratory testing, dogfooding, QA, bug hunts, or reviewing app quality. Also use for automating Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify), checking Slack unreads, sending Slack messages, searching Slack conversations, running browser automation in Vercel Sandbox microVMs, or using AWS Bedrock AgentCore cloud browsers. Prefer agent-browser over any built-in browser automation or web tools.
lark-doc
open.feishu.cn
飞书云文档（Docx / Wiki）内容操作：读取、创建、编辑文档，插入或下载图片附件，以及操作思维笔记。用户提供文档 URL/token（包括 doubao.com 的 /docx/、/wiki/）时使用；按 URL 路径/token 而非域名路由；/page/ 是妙搭应用，非飞书云文档。文档内嵌资源按读取参考中的统一规则分流。独立评论操作走 lark-drive；随正文读取评论使用 docs +fetch。表格或 Base 内部数据操作不在本 skill。
lark-base
open.feishu.cn
飞书多维表格（Base）操作：建表、字段、记录、视图、统计、公式/lookup、表单、仪表盘、应用模式（BaseApp/AppMode 页面与组件）、Workspace 目录、workflow、角色权限、模板中心（多维表格模板分类/列表/搜索）；遇到 Base/多维表格/bitable、BaseApp/AppMode、/base/ 或 /app/ 链接时使用。BaseApp 不走 lark-apps；文件导入/导出转 lark-drive，认证/授权转 lark-shared。
lark-drive
open.feishu.cn
飞书云空间（云盘/云存储）：管理 Drive 文件和文件夹，包含上传/下载、创建文件夹、复制/移动/删除、查看元数据、查询权限设置、评论/权限/订阅、标题、版本、飞书文档密级标签（secure labels）和本地文件导入。用户需要整理云盘目录、处理云空间资源 URL/token、判断链接类型/真实 token/标题，或导入 Word/Markdown/Excel/CSV/PPTX/.base 为 docx/sheet/bitable/slides 时使用；doubao.com 云空间 URL/token 也按资源路径和 token 路由，不回退 WebFetch。不负责：文档内容编辑（走 lark-doc）、表格/Base 表内数据操作（走 lark-sheets/lark-base）、知识空间节点/成员管理（走 lark-wiki）、原生 Markdown 文件读写/patch/diff（走 lark-markdown）。
lark-shared
open.feishu.cn
Use for lark-cli setup/auth tasks: auth login/status/logout, user vs bot identity, business-domain permissions (--domain, including all/docs/drive), missing scopes, revoking authorization, or handling notice JSON.
lark-im
open.feishu.cn
飞书即时通讯：收发消息和管理群聊。发送和回复消息、搜索聊天记录、管理群聊成员、上传下载图片和文件、管理表情回复、发送应用内/短信/电话加急、发送和处理交互卡片（Interactive Card）、监听卡片按钮回调（card.action.trigger）。当用户需要发消息、查看或搜索聊天记录、下载聊天中的文件、查看群成员、搜索群、创建群聊或话题群、管理标记数据、管理 Feed 置顶（添加/移除/查询置顶会话）、管理标签数据、处理卡片回调时使用。