mini-browser — Claude Code skill: a browser for your agent
This skill hands your agent a browser through the mb CLI: open pages, grab screenshots, pull text, fill forms, record screencasts.
npx skills add runablehq/mini-browser --skill mini-browser -a claude-code
npx skills add runablehq/mini-browser --skill skill -a claude-codeWhat mini-browser does
mb is a browser CLI built from small Unix-style commands. It drives Chrome over CDP on port 9222 using puppeteer-core. Your agent runs mb go, mb snap, mb click, mb fill, mb text, mb shot, mb record, mb audit, and reads the result in the terminal or as JSON with --json.
Two things have to be true before anything works: mb is on the machine, and Chrome is listening on the CDP port. Missing pieces are fixed with npm install -g @runablehq/mini-browser and mb-start-chrome. The latter starts Chrome with --remote-debugging-port=9222, a fresh profile and a 1024×768 window, and does nothing if Chrome already runs.
That 1024×768 viewport is where things get annoying. mb snap lists only elements inside the current viewport, so scroll and snap again to see the rest. mb text returns the first querySelector match, which makes mb text "main" a better bet than mb text "p". On heavy SPAs mb go stops at networkidle and that may be too early — follow it with mb wait ".selector".
What it helps with
- open a page and pull its visible text
- fill a form by field labels and submit it
- capture a page as a PNG screenshot
- record an interaction as an mp4 or gif screencast
- run a design audit covering palette, typography, contrast, a11y and SEO
Example prompts for the agent
- Open
https://example.com, snapshot the interactive elements and click the sign-in button - Fill the login form and wait until the URL hits
/dashboard - Record a screencast to
demo.mp4and run the design audit with JSON output
How to install mini-browser
1Skills CLI (Vercel)
npx skills add runablehq/mini-browser --skill mini-browser -a claude-codenpx skills add runablehq/mini-browser --skill skill -a claude-code2Manually: put the files in place
~/.claude/skills/skill/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/skill/
1Skills CLI (Vercel)
npx skills add runablehq/mini-browser --skill skill -a codex2Manually: put the files in place
.agents/skills/skill/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add runablehq/mini-browser --skill skill -a cursor2Manually: put the files in place
.cursor/skills/skill/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install runablehq/mini-browser skill2Manually: put the files in place
.github/skills/skill/SKILL.md1Skills CLI (Vercel)
npx skills add runablehq/mini-browser --skill skill -a gemini-cli2Platform CLI
gemini skills install https://github.com/runablehq/mini-browser --path skillManually: put the files in place
.devin/skills/skill/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add runablehq/mini-browser --skill skill -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about mini-browser
How do I install mini-browser in Claude Code?
Run npx skills add runablehq/mini-browser --skill mini-browser -a claude-code. The skill lands in ~/.claude/skills/skill/SKILL.md.
What does mini-browser do?
It gives your agent browser commands through the mb CLI: navigation, screenshots, text extraction, clicks, form filling, screencast recording, in-page JavaScript and design audits.
What does mini-browser need to run?
mb installed and Chrome listening on CDP port 9222. If the browser is not up, start it with mb-start-chrome.
