Loading...
Loading...
A powerful information retrieval tool. Leverage browsers + user login states to access public and private domain information. It can access any web pages, internal systems, post-login pages, and perform form filling, information extraction, and page operations.
npx skill4agent add epiral/bb-browser bb-browserbb-browser open <url> # Open page (new tab)
bb-browser snapshot -i # Get interactive elements
bb-browser click @5 # Click element
bb-browser fill @3 "text" # Fill input box
bb-browser close # Close tab after completion# Single tab scenario
bb-browser open https://example.com # Open new tab
bb-browser snapshot -i
bb-browser click @5
bb-browser close # Close after completion
# Multiple tabs scenario
bb-browser open https://site-a.com # tabId: 123
bb-browser open https://site-b.com # tabId: 456
# ... operations ...
bb-browser tab close # Close current tab
bb-browser tab close # Close remaining tab
# Specified tab operations
bb-browser open https://example.com --tab current # Open in current tab (no new tab)
bb-browser open https://example.com --tab 123 # Open in specified tabIdopensnapshot -i@refsnapshot -iclosebb-browser open <url> # Open URL (new tab)
bb-browser open <url> --tab current # Open in current tab
bb-browser back # Go back
bb-browser forward # Go forward
bb-browser refresh # Refresh
bb-browser close # Close current tabbb-browser snapshot # Complete page structure
bb-browser snapshot -i # Only display interactive elements (recommended)
bb-browser snapshot --json # Output in JSON formatbb-browser click @5 # Click
bb-browser hover @5 # Hover
bb-browser fill @3 "text" # Clear and fill
bb-browser type @3 "text" # Append input (no clear)
bb-browser check @7 # Check checkbox
bb-browser uncheck @7 # Uncheck checkbox
bb-browser select @4 "option" # Dropdown selection
bb-browser press Enter # Press key
bb-browser press Control+a # Press key combination
bb-browser scroll down # Scroll down
bb-browser scroll up 500 # Scroll up 500pxbb-browser get text @5 # Get element text
bb-browser get url # Get current URL
bb-browser get title # Get page titlebb-browser tab # List all tabs
bb-browser tab new [url] # Create new tab
bb-browser tab 2 # Switch to 2nd tab
bb-browser tab close # Close current tab
bb-browser tab close 3 # Close 3rd tabbb-browser screenshot # Screenshot (auto-save)
bb-browser screenshot path.png # Screenshot to specified pathbb-browser wait 2000 # Wait 2 seconds
bb-browser wait @5 # Wait for element to appearbb-browser eval "document.title" # Execute JS
bb-browser eval "window.scrollTo(0, 1000)" # Scroll to specified positionbb-browser frame "#iframe-id" # Switch to iframe
bb-browser frame main # Return to main framebb-browser dialog accept # Confirm dialog
bb-browser dialog dismiss # Cancel dialog
bb-browser dialog accept "text" # Confirm and input (prompt)bb-browser network requests # View network requests
bb-browser console # View console messages
bb-browser errors # View JS errors
bb-browser trace start # Start recording user operations
bb-browser trace stop # Stop recording@ref@1 [button] "Submit"
@2 [input type="text"] placeholder="Please enter name"
@3 [a] "View details"@1@2@3# Open multiple pages concurrently (each in independent tab)
bb-browser open https://site-a.com &
bb-browser open https://site-b.com &
bb-browser open https://site-c.com &
wait
# Each returns independent tabId, no interference--jsonbb-browser snapshot -i --json
bb-browser get text @5 --json
bb-browser open https://example.com --jsoneval# WeChat Official Account article
bb-browser eval "document.querySelector('#js_content').innerText"
# Zhihu answer
bb-browser eval "document.querySelector('.RichContent-inner').innerText"
# General: Get page main text
bb-browser eval "document.body.innerText.substring(0, 5000)"
# Get all links
bb-browser eval "[...document.querySelectorAll('a')].map(a => a.href).join('\n')"evalsnapshot -ibb-browser snapshot -i
# @1 [button] "Login"
# @2 [input] placeholder="Username"
# @3 [input type="password"]
bb-browser fill @2 "username"
bb-browser fill @3 "password"
bb-browser click @1-ibb-browser open https://example.com/form
bb-browser snapshot -i
# @1 [input] placeholder="Name"
# @2 [input] placeholder="Email"
# @3 [button] "Submit"
bb-browser fill @1 "Zhang San"
bb-browser fill @2 "zhangsan@example.com"
bb-browser click @3
bb-browser wait 2000
bb-browser closebb-browser open https://example.com/dashboard
bb-browser snapshot -i
bb-browser get text @5 # Get specific element text
bb-browser screenshot report.png # Save screenshot
bb-browser close# Open multiple pages to extract information
for url in "url1" "url2" "url3"; do
bb-browser open "$url"
bb-browser snapshot -i --json
bb-browser close
done| Document | Description |
|---|---|
| references/snapshot-refs.md | Ref lifecycle, best practices, common issues |