Xiaopi Agent Browser

A fast Rust-based headless browser automation CLI with Node.js fallback that enables AI agents to navigate, click, type, and snapshot pages via structured co...

Security Scan

The skill appears to be an instruction-only wrapper for an agent-browser CLI (browser automation) but contains inconsistent metadata and install expectations (missing source/homepage, owner/version mismatches, and contradictory binary requirements), so verify the upstream package and contents before installing or allowing autonomous use.

Purpose & Capability

The SKILL.md describes a Rust-based CLI with a Node.js fallback and shows npm-based installation instructions. The registry metadata requires node and npm as binaries, yet a Rust binary would not normally require npm. The SKILL.md also documents building from source (git + pnpm) but the skill's declared required binaries do not include git or pnpm. Additionally, the skill registry entry (slug/version/owner) does not match the _meta.json contents (different ownerId, slug, and version), and the package has no declared homepage or source URL in the registry — these inconsistencies suggest packaging/copying or provenance issues.

Instruction Scope

The runtime instructions are narrowly focused on browser automation commands (navigate, snapshot, click, fill, upload, screenshot, etc.), which aligns with the described purpose. The document includes commands that can access pages, cookies/storage, and upload local files — expected for a browser automation tool but potentially capable of exposing sensitive local data if misused. The SKILL.md does not instruct reading unrelated system files or environment variables.

Install Mechanism

There is no formal install spec in the skill bundle (instruction-only), but SKILL.md recommends npm global installation (npm install -g agent-browser) and provides a from-source path requiring git and pnpm. Running npm install -g executes package install scripts from the npm registry, which can run arbitrary code on the host. Because the registry metadata has no source/homepage and the skill manifest mismatches the included _meta.json, it's unclear whether the npm package name and upstream repository are trustworthy. The absence of declared git/pnpm in required binaries is another mismatch.

Credentials

The skill declares no required environment variables or credentials, and SKILL.md does not request secrets. That is proportionate for a browser automation CLI. However, the tool's capabilities (navigation, cookies, file upload) mean an installed CLI could access local files and network endpoints, so installation should be treated with the same caution as installing any third-party CLI.

Persistence & Privilege

The skill does not request always:true and uses default agent invocation settings. It is instruction-only and does not include code that would be written to disk by the platform. Autonomous invocation is allowed (platform default) — combine that with the install concerns above when deciding whether to permit autonomous runs.

What to consider before installing

This skill is plausibly a useful browser-automation wrapper, but there are red flags around provenance and install instructions. Before installing or allowing the agent to run it autonomously: 1) Verify the npm package name 'agent-browser' and inspect the package on the npm registry for publisher, version, and install scripts; 2) Confirm the upstream repository (the SKILL.md references github.com/vercel-labs/agent-browser) — check the repo owner, tags, and source code; 3) Resolve metadata mismatches (ownerId/slug/version differences between registry metadata and _meta.json) with the publisher; 4) If you must test, install and run it in a sandbox or isolated VM/container and avoid global npm installs on production hosts; 5) Limit the agent's access to local files and outgoing network when first enabling the skill. If you cannot verify the upstream package and author, treat the skill as untrusted.

Browser Automation with agent-browser

Installation

npm recommended

npm install -g agent-browser
agent-browser install
agent-browser install --with-deps

From Source

git clone https://github.com/vercel-labs/agent-browser
cd agent-browser
pnpm install
pnpm build
agent-browser install

Quick start

agent-browser open <url>        # Navigate to page
agent-browser snapshot -i       # Get interactive elements with refs
agent-browser click @e1         # Click element by ref
agent-browser fill @e2 "text"   # Fill input by ref
agent-browser close             # Close browser

Core workflow

1. Navigate: agent-browser open <url>
2. Snapshot: agent-browser snapshot -i (returns elements with refs like @e1, @e2)
3. Interact using refs from the snapshot
4. Re-snapshot after navigation or significant DOM changes

Commands

Navigation

agent-browser open <url>      # Navigate to URL
agent-browser back            # Go back
agent-browser forward         # Go forward
agent-browser reload          # Reload page
agent-browser close           # Close browser

Snapshot (page analysis)

agent-browser snapshot            # Full accessibility tree
agent-browser snapshot -i         # Interactive elements only (recommended)
agent-browser snapshot -c         # Compact output
agent-browser snapshot -d 3       # Limit depth to 3
agent-browser snapshot -s "#main" # Scope to CSS selector

Interactions (use @refs from snapshot)

agent-browser click @e1           # Click
agent-browser dblclick @e1        # Double-click
agent-browser focus @e1           # Focus element
agent-browser fill @e2 "text"     # Clear and type
agent-browser type @e2 "text"     # Type without clearing
agent-browser press Enter         # Press key
agent-browser press Control+a     # Key combination
agent-browser keydown Shift       # Hold key down
agent-browser keyup Shift         # Release key
agent-browser hover @e1           # Hover
agent-browser check @e1           # Check checkbox
agent-browser uncheck @e1         # Uncheck checkbox
agent-browser select @e1 "value"  # Select dropdown
agent-browser scroll down 500     # Scroll page
agent-browser scrollintoview @e1  # Scroll element into view
agent-browser drag @e1 @e2        # Drag and drop
agent-browser upload @e1 file.pdf # Upload files

Get information

agent-browser get text @e1        # Get element text
agent-browser get html @e1        # Get innerHTML
agent-browser get value @e1       # Get input value
agent-browser get attr @e1 href   # Get attribute
agent-browser get title           # Get page title
agent-browser get url             # Get current URL
agent-browser get count ".item"   # Count matching elements
agent-browser get box @e1         # Get bounding box

Check state

agent-browser is visible @e1      # Check if visible
agent-browser is enabled @e1      # Check if enabled
agent-browser is checked @e1      # Check if checked

Screenshots & PDF

agent-browser screenshot          # Screenshot to stdout
agent-browser screenshot path.png # Save to file
agent-browser screenshot --full   # Full page
agent-browser pdf output.pdf      # Save as PDF

Video recording

agent-browser record start ./demo.webm    # Start recording (uses current URL + state)
agent-browser click @e1                   # Perform actions
agent-browser record stop                 # Stop and save video
agent-browser record restart ./take2.webm # Stop current + start new recording

Recording creates a fresh context but preserves cookies/storage from your session. If no URL is provided, it automatically returns to your current page. For smooth demos, explore first, then start recording.

Wait

agent-browser wait @e1                     # Wait for element
agent-browser wait 2000                    # Wait milliseconds
agent-browser wait --text "Success"        # Wait for text
agent-browser wait --url "/dashboard"    # Wait for URL pattern
agent-browser wait --load networkidle      # Wait for network idle
agent-browser wait --fn "window.ready"     # Wait for JS condition

Mouse control

agent-browser mouse move 100 200      # Move mouse
agent-browser mouse down left         # Press button
agent-browser mouse up left           # Release button
agent-browser mouse wheel 100         # Scroll wheel

Semantic locators (alternative to refs)

agent-browser find role button click --name "Submit"
agent-browser find text "Sign In" click
agent-browser find label "Email" fill "[e-mail]"
agent-browser find first ".item" click
agent-browser find nth 2 "a" text

Browser settings

agent-browser set viewport 1920 1080      # Set viewport size
agent-browser set device "iPhone 14"      # Emulate device
agent-browser set geo 37.7749 -122.4194   # Set geolocation
agent-browser set offline on              # Toggle offline mode
agent-browser set headers '{"X-Key":"v"}' # Extra HTTP headers
agent-browser set credentials user pass   # HTTP basic auth
agent-browser set media dark              # Emulate color scheme

Cookies & Storage

agent-browser cookies                     # Get all cookies
agent-browser cookies set name value      # Set cookie
agent-browser cookies clear               # Clear cookies
agent-browser storage local               # Get all localStorage
agent-browser storage local key           # Get specific key
agent-browser storage local set k v       # Set value
agent-browser storage local clear         # Clear all

Network

agent-browser network route <url>              # Intercept requests
agent-browser network route <url> --abort      # Block requests
agent-browser network route <url> --body '{}'  # Mock response
agent-browser network unroute [url]            # Remove routes
agent-browser network requests                 # View tracked requests
agent-browser network requests --filter api    # Filter requests

Tabs & Windows

agent-browser tab                 # List tabs
agent-browser tab new [url]       # New tab
agent-browser tab 2               # Switch to tab
agent-browser tab close           # Close tab
agent-browser window new          # New window

Frames

agent-browser frame "#iframe"     # Switch to iframe
agent-browser frame main          # Back to main frame

Dialogs

agent-browser dialog accept [text]  # Accept dialog
agent-browser dialog dismiss        # Dismiss dialog

JavaScript

agent-browser eval "document.title"   # Run JavaScript

State management

agent-browser state save auth.json    # Save session state
agent-browser state load auth.json    # Load saved state

Example: Form submission

agent-browser open https://example.com/form
agent-browser snapshot -i
# Output shows: textbox "Email" [ref=e1], textbox "Password" [ref=e2], button "Submit" [ref=e3]

agent-browser fill @e1 "[e-mail]"
agent-browser fill @e2 "password123"
agent-browser click @e3
agent-browser wait --load networkidle
agent-browser snapshot -i  # Check result

Example: Authentication with saved state

# Login once
agent-browser open https://app.example.com/login
agent-browser snapshot -i
agent-browser fill @e1 "username"
agent-browser fill @e2 "password"
agent-browser click @e3
agent-browser wait --url "/dashboard"
agent-browser state save auth.json

# Later sessions: load saved state
agent-browser state load auth.json
agent-browser open https://app.example.com/dashboard

Sessions (parallel browsers)

agent-browser --session test1 open site-a.com
agent-browser --session test2 open site-b.com
agent-browser session list

JSON output (for parsing)

Add --json for machine-readable output:

agent-browser snapshot -i --json
agent-browser get text @e1 --json

Debugging

agent-browser open example.com --headed              # Show browser window
agent-browser console                                # View console messages
agent-browser console --clear                        # Clear console
agent-browser errors                                 # View page errors
agent-browser errors --clear                         # Clear errors
agent-browser highlight @e1                          # Highlight element
agent-browser trace start                            # Start recording trace
agent-browser trace stop trace.zip                   # Stop and save trace
agent-browser record start ./debug.webm              # Record from current page
agent-browser record stop                            # Save recording
agent-browser --cdp 9222 snapshot                    # Connect via CDP

Troubleshooting

• If the command is not found on Linux ARM64, use the full path in the bin folder.
• If an element is not found, use snapshot to find the correct ref.
• If the page is not loaded, add a wait command after navigation.
• Use --headed to see the browser window for debugging.

Options

• --session <name> uses an isolated session.
• --json provides JSON output.
• --full takes a full page screenshot.
• --headed shows the browser window.
• --timeout sets the command timeout in milliseconds.
• --cdp <port> connects via Chrome DevTools Protocol.

Notes

• Refs are stable per page load but change on navigation.
• Always snapshot after navigation to get new refs.
• Use fill instead of type for input fields to ensure existing text is cleared.

Reporting Issues

• Skill issues: Open an issue at https://github.com/TheSethRose/Agent-Browser-CLI
• agent-browser CLI issues: Open an issue at https://github.com/vercel-labs/agent-browser