Skills browser-use
🌐

browser-use

Low Risk ⚙️ External commands🌐 Network access📁 Filesystem access

Automate browser interactions and web tasks

Also available from: AbdullahMalik17,92Bilal26,92Bilal26

Manual web testing and data extraction is time-consuming and error-prone. Browser-use provides AI-driven browser automation for reliable, repeatable web interactions.

Supports: Claude Codex Code(CC)
⚠️ 67 Poor
1

Download the skill ZIP

2

Upload in Claude

Go to Settings → Capabilities → Skills → Upload skill

3

Toggle on and start using

Test it

Using "browser-use". browser-use state on a search results page

Expected outcome:

Page state showing URL, title, and numbered list of clickable elements like '1. Search result: Article Title', '2. Next page button'

Using "browser-use". browser-use screenshot with save path

Expected outcome:

Screenshot saved to specified file path with confirmation message and base64 preview

Security Audit

Low Risk
v2 • 3/19/2026

Static analysis flagged 163 patterns in SKILL.md documentation file. All findings are false positives: backtick patterns are markdown code blocks documenting CLI usage, URLs are example/documentation links, and file paths reference user-controlled locations for legitimate browser profile and cookie management. The skill provides browser automation via the browser-use CLI with no malicious intent detected.

1
Files scanned
547
Lines analyzed
6
findings
2
Total audits
Medium Risk Issues (1)
SKILL.md:405-407
Cookie Export/Import Capability
The skill supports exporting cookies to files and importing them, which could expose session tokens if misused. This is legitimate browser profile syncing functionality but should be documented.
Low Risk Issues (2)
SKILL.md:45-80
External Command Documentation
SKILL.md contains extensive documentation of browser-use CLI commands. All instances are markdown code blocks for documentation purposes, not actual code execution.
SKILL.md:19
Hardcoded URLs in Documentation
Multiple example URLs present in SKILL.md including github.com, example.com. These are documentation links and example values, not hardcoded endpoints for data exfiltration.

Risk Factors

⚙️ External commands (2)
SKILL.md:15-17 SKILL.md:45-80
🌐 Network access (2)
SKILL.md:19 SKILL.md:204
📁 Filesystem access (2)
SKILL.md:40 SKILL.md:405-407
Audited by: claude View Audit History →

Quality Score

38
Architecture
100
Maintainability
87
Content
27
Community
81
Security
87
Spec Compliance

What You Can Build

Web QA Testing

Automate end-to-end testing workflows: navigate to staging site, verify elements, fill forms, submit, and validate results with screenshots.

Authenticated Data Extraction

Sync browser profiles to access authenticated sites, extract data from dashboards or reports, and export results without manual login.

Autonomous Research Agents

Deploy cloud browser agents to research topics across multiple sites, summarize findings, and run parallel research tasks.

Try These Prompts

Basic Page Navigation 
Open https://example.com and show me the available clickable elements.
Form Filling Workflow 
Navigate to the login page, check the available elements, input username into field 3 and password into field 5, then click the submit button.
Screenshot and Data Extraction 
Take a screenshot of the current page, get the page title, extract all text from element 12, and execute JavaScript to get the current URL.
Cloud Agent Research Task 
Run an autonomous agent task: Search for the latest AI news on three major tech sites and summarize the top 5 stories with links.

Best Practices

  • Always run 'browser-use state' first to see available elements and their indices before interacting
  • Use domain-specific cookie sync instead of full profile sync to minimize exposure of sensitive data
  • Close browser sessions and stop cloud sessions after completing tasks to free resources

Avoid

  • Do not use full cookie profile sync without explicit user consent - always specify domains
  • Do not reuse sessions after stopping tasks - create new sessions instead
  • Do not run cloud agent tasks without monitoring - check status periodically and stop stuck tasks

Frequently Asked Questions

What is browser-use and how do I install it?
Browser-use is a CLI tool for browser automation. Run 'browser-use doctor' to check installation status and see setup instructions.
How do I use this skill with my existing browser login sessions?
Use '--browser real --profile ProfileName' for local Chrome profiles, or sync cookies to a cloud profile using 'browser-use profile sync --from ProfileName --domain example.com'.
What is the difference between chromium, real, and remote browser modes?
Chromium is fast and headless by default. Real uses your local Chrome with existing logins. Remote runs in the cloud with proxy support and AI agent capabilities.
How do I run autonomous AI agent tasks?
Use 'browser-use -b remote run task description' to launch an AI agent task. Poll status with 'browser-use task status task-id' and stop with 'browser-use task stop task-id'.
Can I expose my local dev server to a cloud browser?
Yes, use 'browser-use tunnel port' to create a Cloudflare tunnel URL, then open that URL with '--browser remote'.
How do I clean up browser sessions and resources?
Run 'browser-use close' for local sessions, 'browser-use session stop --all' for cloud sessions, and 'browser-use tunnel stop --all' for tunnels.

Developer Details

License

MIT

Repository

https://github.com/browser-use/browser-use/tree/main/skills/browser-use/

Ref

main

File structure

📄 SKILL.md