agentsclimarketplace

Stealth browser

Skill zippoxer/claude-skills/skills/stealth-browser

Stealth browser automation with anti-bot bypass and persistent page state. Use when users need to navigate Cloudflare-protected sites, fill forms, take screenshots, extract web data, or automate browser workflows on sites that block regular automation. Trigger phrases include "go to [url]", "click on", "fill out the form", "take a screenshot", "scrape", "automate", "bypass cloudflare", or any browser interaction request on protected sites.From its SKILL.md

Install
npx -y skills add zippoxer/claude-skills --skill stealth-browser

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 5 stars5 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

6.4 KB, as published. Nobody here has run it

Stealth Browser Skill

Browser automation using nodriver for anti-bot/Cloudflare bypass. Maintains page state across script executions. Write small, focused scripts to accomplish tasks incrementally.

Key difference from dev-browser: Uses nodriver instead of Playwright, which bypasses most anti-bot detection including Cloudflare Turnstile.

Choosing Your Approach

  • Local/source-available sites: Read the source code first to write selectors directly
  • Unknown page layouts: Use get_ai_snapshot() to discover elements and select_snapshot_ref() to interact with them
  • Visual feedback: Take screenshots to see what the user sees

Setup

Important: Each session needs its own server. Do not reuse servers started by other sessions.

Start your server:

./skills/stealth-browser/server.sh &
  • Default port is 6222. If in use, the script will tell you which port to use instead.
  • Add --headless flag if user requests headless mode.
  • Wait for the Ready on port XXXX message before running scripts.
  • Remember the port number - you'll need it for connect(port=XXXX).

If you see "Port 6222 is in use", that's another session's server. Start your own on the suggested port:

./skills/stealth-browser/server.sh --port 6223 &

Writing Scripts

Execute scripts inline using heredocs. Use the port from your server startup:

cd skills/stealth-browser && source venv/bin/activate && python <<'EOF'
from client import connect

client = connect(port=6222)  # Use YOUR server's port!
page = client.page("example")  # descriptive name like "cnn-homepage"

page.goto("https://example.com")

print(f"Title: {page.title()}")
print(f"URL: {page.url}")

client.disconnect()
EOF

Write to tmp/ files only when the script needs reuse, is complex, or user explicitly requests it.

Key Principles

  1. Small scripts: Each script does ONE thing (navigate, click, fill, check)
  2. Evaluate state: Print state at the end to decide next steps
  3. Descriptive page names: Use "checkout", "login", not "main"
  4. Disconnect to exit: client.disconnect() - pages persist on server
  5. Plain JS in evaluate: page.evaluate() runs in browser - use JavaScript syntax

Workflow Loop

Follow this pattern for complex tasks:

  1. Write a script to perform one action
  2. Run it and observe the output
  3. Evaluate - did it work? What's the current state?
  4. Decide - is the task complete or do we need another script?
  5. Repeat until task is done

Client API

from client import connect

client = connect(port=6222)                 # Connect to YOUR server's port!
page = client.page("name")                  # Get or create named page
pages = client.list()                       # List all page names
client.close("name")                        # Close a page
client.disconnect()                         # Disconnect (pages persist)

# ARIA Snapshot methods
snapshot = client.get_ai_snapshot("name")   # Get accessibility tree (YAML)
ref = client.select_snapshot_ref("name", "e5")  # Get element proxy by ref
ref.click()                                 # Click the ref
ref.fill("text")                            # Fill the ref with text

Page Operations

client = connect(port=6222)  # Your server's port
page = client.page("mypage")

# Navigation
page.goto("https://example.com")

# Get state
url = page.url                    # Current URL (property)
title = page.title()              # Page title

# Interaction
page.click("button.submit")       # Click by CSS selector
page.fill("input[name=q]", "search term")  # Fill input

# JavaScript
result = page.evaluate("document.title")   # Run JS, get result

# Screenshots
page.screenshot("screenshot.png")          # Saves to tmp/ directory

# Waiting
page.wait_for_selector(".results")         # Wait for element

Inspecting Page State

Screenshots

page.screenshot("screenshot.png")      # Saves to tmp/screenshot.png
page.screenshot("tmp/full.png")        # Explicit path

ARIA Snapshot (Element Discovery)

Use get_ai_snapshot() to discover page elements. Returns YAML-formatted accessibility tree:

- banner:
  - link "Hacker News" [ref=e1]
  - navigation:
    - link "new" [ref=e2]
- main:
  - list:
    - listitem:
      - link "Article Title" [ref=e8]
      - link "328 comments" [ref=e9]
- contentinfo:
  - textbox [ref=e10]
    - /placeholder: "Search"

Interpreting refs:

  • [ref=eN] - Element reference for interaction (visible, clickable elements only)
  • [checked], [disabled], [expanded] - Element states
  • [level=N] - Heading level
  • /url:, /placeholder: - Element properties

Interacting with refs:

snapshot = client.get_ai_snapshot("hackernews")
print(snapshot)  # Find the ref you need

element = client.select_snapshot_ref("hackernews", "e2")
element.click()

Error Recovery

Page state persists after failures. Debug with:

cd skills/stealth-browser && source venv/bin/activate && python <<'EOF'
from client import connect

client = connect(port=6222)  # Your server's port
page = client.page("debug")

page.screenshot("debug.png")
print(f"URL: {page.url}")
print(f"Title: {page.title()}")
print(f"Body: {page.evaluate('document.body.innerText.substring(0, 200)')}")

client.disconnect()
EOF

Cloudflare Bypass

This skill uses nodriver which has built-in stealth patches. Sites protected by Cloudflare (like lowendtalk.com) work automatically:

page.goto("https://lowendtalk.com/")
# No Cloudflare challenge - loads directly!
print(page.title())  # "LowEndTalk" (not "Just a moment...")

Differences from dev-browser

Aspectdev-browserstealth-browser
LanguageTypeScriptPython
Browser libPlaywrightnodriver (stealth)
Anti-botNoneBuilt-in bypass
Port92226222
Extension modeYesNo
Script syntaxnpx tsx <<'EOF'python <<'EOF'

Keep looking

Skills are one crate of 610,220. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.