skill CX Codex

agent-browser · 101-skills — Claude Code skill: a browser for your agent

Q: How do I install agent-browser in Claude Code?

Run `npx skills add 101-skills/superpowers --skill agent-browser -a claude-code`. The skill lands at `~/.claude/skills/agent-browser/SKILL.md`.

Q: What do I need before using it?

The `belt` CLI from inference.sh plus a `belt login` session. Without those, the skill's commands won't run.

This skill hands your agent a browser: open pages, click elements, fill forms, grab screenshots and record video. Useful for scraping and UI testing.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓

npx skills add 101-skills/superpowers --skill agent-browser -a claude-code

Open on GitHub

What agent-browser does

The skill runs on inference.sh and its belt CLI. Your agent opens a URL with belt app run agent-browser --function open and gets back a list of interactive elements tagged @e1, @e2 and so on. From there it clicks, types, selects, drags or uploads files by referencing those tags. Playwright does the actual work underneath, but the agent never touches selectors.

Those refs go stale fast. Any navigation, form submit or dynamic content load invalidates them, so you have to call snapshot again — otherwise the agent keeps clicking numbers that no longer point anywhere. A session persists across calls; open returns its id and close ends it, handing back the video file if recording was on.

Before anything runs you need the CLI installed and a belt login session. The function set covers JavaScript execution on the page, screenshots, proxy routing and video capture with a cursor indicator. Bundled templates handle form filling, authenticated sessions and page capture.

What it helps with

fill out a web form and submit it
capture a viewport or full-page screenshot
pull data off a page with custom JavaScript
record a walkthrough video for debugging
upload a file into a page input

Example prompts for the agent

Open https://example.com/login and list the interactive elements
Fill the Email field with [email protected] and click the sign-in button
Take a full-page screenshot and record the session as video

How to install agent-browser

1Skills CLI (Vercel)

npx skills add 101-skills/superpowers --skill agent-browser -a claude-code

2Manually: put the files in place

~/.claude/skills/agent-browser/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/agent-browser/

1Skills CLI (Vercel)

npx skills add 101-skills/superpowers --skill agent-browser -a codex

2Manually: put the files in place

.agents/skills/agent-browser/SKILL.md

For all projects use ~/.agents/skills/

1Skills CLI (Vercel)

npx skills add 101-skills/superpowers --skill agent-browser -a cursor

2Manually: put the files in place

.cursor/skills/agent-browser/SKILL.md

Cursor also reads .agents/skills/ and .claude/skills/

1GitHub CLI

gh skill install 101-skills/superpowers agent-browser

2Manually: put the files in place

.github/skills/agent-browser/SKILL.md

1Skills CLI (Vercel)

npx skills add 101-skills/superpowers --skill agent-browser -a gemini-cli

2Platform CLI

gemini skills install https://github.com/101-skills/superpowers --path tools/utilities/agent-browser

Manually: put the files in place

.devin/skills/agent-browser/SKILL.md

The legacy .windsurf/skills/ path still works

Skills CLI (Vercel)

npx skills add 101-skills/superpowers --skill agent-browser -a opencode

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about agent-browser

How do I install agent-browser in Claude Code?

Run npx skills add 101-skills/superpowers --skill agent-browser -a claude-code. The skill lands at ~/.claude/skills/agent-browser/SKILL.md.

What does agent-browser do?

It lets an agent drive a browser: navigate, click, type, upload files, take screenshots, record video and run JavaScript on the page.

What do I need before using it?

The belt CLI from inference.sh plus a belt login session. Without those, the skill's commands won't run.

Similar skills

All skills for Browser and scraping →

agent-browser by other authors comes first.

agent-browser vercel-labs A CLI skill for browser control: it opens pages, fills forms, clicks buttons, and tests web apps. Use it when an agent needs to act inside a website. CX Codex +1 1,200,000installs agent-browser nexu-io Automates browser actions via CLI and CDP. Lets AI agents inspect local web apps, run tests, take element screenshots, and extract design tokens. CX Codex +1 2,800installs agent-browser official Vercel Labs Automates browser actions via CLI, takes DOM snapshots, clicks elements, and verifies dev servers. Helps AI agents inspect web interfaces directly. CX Codex GP ChatGPT 20installs webapp-testing official anthropics This skill tests local web apps with Playwright, helping agents debug UI behavior, capture screenshots, and inspect browser logs. 186,758installs computer-use stablyai This skill drives a visible app window through orca computer: accessibility tree, clicks, typing, screenshots. Reach for it when the job sits behind a GUI. CX Codex +1 239,400installs browser-use browser-use Hands the agent a real Chrome over CDP: clicks, typing, screenshots, and pages that need JS or your logged-in session. CX Codex +1 98,200installs playwright-cli microsoft A skill that hands your agent a browser: open a page, click elements, grab a snapshot, and work with Playwright tests. CX Codex +1 195,700installs last30days mvanhorn Pulls posts and engagement on any topic from the last 30 days across Reddit, X, YouTube, TikTok, Hacker News, Polymarket, GitHub and the web. Useful when you want what people actually said, not what a blog summarized. CX Codex +1 53,000installs

From the same set22

ai-video-generation Generate videos with Google Veo, Seedance, HappyHorse and 40+ models directly in your editor. Supports text-to-video, image animation, lipsync and audio generation. CX Codex +1 872,100installs ai-image-generation Generates and edits images through belt with selectable models. Use it for illustrations, mockups, and changes to existing images. CX Codex +1 870,900installs twitter-automation Automates Twitter and X interactions through the belt command-line interface. It posts content, manages direct messages, tracks profiles, and handles likes. CX Codex +1 870,600installs ai-avatar-video A skill for making AI avatar and talking-head videos through the inference.sh CLI. Useful for presenter clips, UGC ads and dubbing. CX Codex +1 860,700installs web-search This skill gives your agent web search and content extraction through Tavily and Exa. Useful for research, fact-checking and RAG pipelines. CX Codex +1 110,000installs python-executor Runs Python 3.10 scripts inside an isolated inference.sh sandbox. Helpful when your agent needs to crunch data, scrape websites, or render graphics without touching your local system. CX Codex +1 109,700installs remotion-render This skill sends a Remotion React component to inference.sh and gets back an MP4. Useful when video should come from code, not from a timeline editor. CX Codex +1 109,700installs landing-page-design Guides landing page work around the hero, CTA and social proof. Use it when shaping a product or marketing page and its visual direction. CX Codex +1 56,100installs