agent-browser

9716 stars 508 forks

Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.

Marketplace

#Automates browser#references#web testing#templates#Automates#web#form filling#form

Third-Party Agent Skill: Review the code before installing. Agent skills execute in your AI assistant's environment and can access your files. Learn more about security

Installation for Agentic Skill

View all platforms →

Claude Code (CLI) Fast

skilz install vercel-labs/agent-browser/agent-browser

OpenCode (CLI) Fast

skilz install vercel-labs/agent-browser/agent-browser --agent opencode

OpenAI Codex (CLI) Native

skilz install vercel-labs/agent-browser/agent-browser --agent codex

Gemini CLI (Project) Project

skilz install vercel-labs/agent-browser/agent-browser --agent gemini

First time? Install Skilz: pip install skilz

Works with 22+ AI coding assistants

Cursor, Aider, Copilot, Windsurf, Qwen, Kimi, and more...

View All Agents

For Claude Desktop Easy

Download Agent Skill ZIP

Extract and copy to ~/.claude/skills/ then restart Claude Desktop

Manual Installation

1. Clone the repository:

git clone https://github.com/vercel-labs/agent-browser

2. Copy the agent skill directory:

cp -r agent-browser/skills/agent-browser ~/.claude/skills/

Agentic Skill Grades →

Agent Skill Grade

Score: 66/100 Click to see breakdown

Score Breakdown

Spec Compliance

12/15

PDA Architecture

20/30

Ease of Use

17/25

Writing Style

8/10

Utility

11/20

Modifiers: -2

Recommendations

Focus on improving Utility (currently 11/20)
Add trigger phrases to description for discoverability
Add table of contents for files over 100 lines

Graded: 2026-01-24

Developer Feedback

I was curious about your browser automation approach, but the missing README made it hard to understand the actual use cases—are you targeting web testing, data extraction, or something else entirely?

Links:

TL;DR

You're at 66/100, solidly in D territory—needs work before this is production-ready. The evaluation uses Anthropic's skill best practices, and your strongest area is Spec Compliance (12/15), which means your YAML frontmatter is solid. The weak spot is Utility (11/20) by a fair margin—your skill does the plumbing but doesn't show enough of why someone would use it or what they can actually do with it.

What's Working Well

Clean frontmatter: Your YAML metadata is valid and hits all required fields—no errors there
Proper naming: agent-browser follows hyphen-case conventions correctly
Structured references: You've got 5 reference files (authentication, proxy-support, video-recording, snapshot-refs, session-management) which shows you're thinking about layered documentation
Good examples: You've got code blocks sprinkled through, which nets a +2 bonus for exemplary examples

The Big One: Utility is Your Blocker (11/20 → Target 18+)

This is the killer. Your skill describes what it does but barely shows why I'd use it or what I can build with it. You've got 1/4 points for feedback loops and only 5/8 for problem-solving power.

The fix: Add concrete use-case sections to your main SKILL.md. Right now it's too API-reference-heavy. You need something like:

## Common Use Cases

- **Automated testing**: Verify UI behavior across browsers
- **Data extraction**: Scrape dynamic content that requires JavaScript
- **Screenshot monitoring**: Compare visual changes over...

AI-Detected Topics

Extracted using NLP analysis

Automates browser references web testing templates Automates web form filling form screenshots web pages

Report Security Issue

Found a security vulnerability in this agent skill?