A collection of Browser Skills for OPENCLAW-style agents and Skill workflows — built for runs that break on verification, redirects, unstable sessions, or actions that stop working mid-flow.
- The task starts fine — then Cloudflare appears and everything stops
- The page loads, but clicks stop working halfway through
- Login works, then the session quietly dies on the next redirect
- The agent runs for minutes, then returns nothing. Sound familiar?
If any of these have killed your workflow, BrowserAct is for you.
BrowserAct ships two flagship products. Pick the one that fits your problem — or use them together.
Drives a real, stealth-capable Chrome from the command line. Your agent navigates, clicks, fills forms, handles captchas, and operates websites just like a human would.
npx skills add browser-act/skills --skill browser-actA meta-Skill that explores a site once, then generates a deploy-ready Skill package (SKILL.md + Python scripts). Every subsequent call skips the exploration cost.
npx skills add browser-act/skills --skill browser-act-skill-forgeUse browser-act CLI when... |
Use Skill Forge when... |
|---|---|
| You want the agent to operate a website live | You need to extract or act repeatedly at scale |
| One-off tasks, ad-hoc browsing | Batch jobs, recurring workflows, zero-maintenance automation |
| Login flows, captchas, multi-account demos | Hundreds/thousands of records, cross-pagination, recurring monitoring |
Fast, persistent browser automation from the command line:
browser-act navigate https://example.com # Navigate to URL
browser-act state # See clickable elements
browser-act click 5 # Click element by index
browser-act input 3 "Hello" # Click and type text
browser-act screenshot page.png # Take screenshotIt supports two browser types (Stealth and Real Chrome) and provides scriptable commands for navigation, page interaction, data extraction, and session management.
Most AI agents get blocked by Cloudflare, CAPTCHAs, or login walls. browser-act provides the missing infrastructure to let your agent browse like a real human.
| 🛡️ Anti-Detection Stealth | Bypasses Cloudflare, reCAPTCHA, Datadome, and more. Authentic browser fingerprints — your Agent won't get blocked. |
| 🔗 Real Chrome Control | Drive your existing Chrome with all logins, cookies, and extensions. No re-authentication needed. |
| 🚀 Parallel Execution | Run multiple stealth browsers concurrently — each with independent fingerprints, proxies, and sessions. |
| 🤖 Captcha Solving | Built-in automatic captcha solving. No third-party services or manual intervention required. |
| 📉 Low Token Noise | Strips 90% of junk HTML before feeding to the LLM. Save money, get faster answers. |
| 🔒 Proxy & Privacy Modes | Per-browser proxy support (HTTP/SOCKS5) and privacy mode for fresh environments on every launch. |
These aren't features for edge cases. They're fixes for the exact failures that stop most agents cold.
Every time you ask an AI agent to batch-extract from a new website, it starts from scratch — different path each time, different failure modes, unreliable at scale. If the site changes layout, the whole thing breaks.
Skill Forge fixes that: explore once, reuse forever.
How it works:
- Describe — tell the agent the target site and what data to extract or what action to perform
- Explore — Skill Forge automatically discovers API endpoints (preferred) or DOM patterns
- Generate — produces deploy-ready Skill files with business parameters extracted as CLI arguments
- Auto-test — a sub-agent verifies the Skill end-to-end and self-fixes failures until passing
- Deploy & Scale — run 500 or 5,000 records through the same stable path, anytime
npx skills add browser-act/skills --skill browser-actThen tell your agent:
"Go to Amazon.com, find the top 10 best-selling products in the mouse category, and save them to a markdown file."
npx skills add browser-act/skills --skill browser-act-skill-forgeThen tell your agent:
"Forge a Skill that extracts job listings from LinkedIn — title, company, salary, URL. I'll run it on 300 keywords later."
Your agent will prompt you to register and get a free API Key interactively if it encounters advanced anti-bot protections.
🚀 Works Seamlessly Across All Major AI Assistants
BrowserAct skills are designed to work powerfully and reliably on all leading AI coding platforms:
| Platform | Status | Installation |
|---|---|---|
| Claude Code | ✅ Fully Supported | Native skill support |
| Cursor | ✅ Fully Supported | Works out of the box |
| VS Code | ✅ Fully Supported | Works out of the box |
| OpenCode | ✅ Fully Supported | Direct integration |
| OpenClaw | ✅ Fully Supported | Compatible |
| Codex | ✅ Fully Supported | Compatible |
| Gemini CLI | ✅ Fully Supported | Compatible |
Key Benefits:
- ✅ Stable & Reliable: No crashes, no unexpected behavior.
- ✅ Plug & Play: Works immediately after installation.
- ✅ Cross-Platform: Consistent performance across all AI assistants.
This repository contains ready-to-use scenario skills covering E-commerce scraping, social media monitoring, lead generation, and more.
Here are some of the highlighted skills you can use right away:
- Amazon ASIN Lookup Skill: Extract structured product details from Amazon using ASIN.
- Amazon Best Selling Products Finder: Extract best-selling product data based on keywords.
- Google News API Skill: Track industry trends and breaking news.
- Google Maps API Skill: Scrape local business data and contact info.
- YouTube Transcript Extractor: Extract transcripts and metadata from YouTube videos automatically.
Browse all available Skills in the solutions directory →
Works with any scenario. Can't find what you need? Join Discord and request a Skill.
BrowserAct Skills is free and open source. If it saves you time, please give us a ⭐ Star — it keeps the project alive and helps us ship more skills.
🎁 Bonus: Once you star the repository, you can join our Discord and post in the #claim-500-credits channel to receive 500 free credits!
Built with ❤️ by the BrowserAct Team