browser-automation（浏览器自动化）

Browser Automation

Automate browser interactions using Stagehand CLI with Claude. This skill provides natural language control over a Chrome browser through command-line tools for navigation, interaction, data extraction, and screenshots.

Overview

This skill uses a CLI-based approach where Claude Code calls browser automation commands via bash. The browser stays open between commands for faster sequential operations and preserves browser state (cookies, sessions, etc.).

Setup Verification

IMPORTANT: Before using any browser commands, you MUST check setup.json in this directory.

First-Time Setup Check

1. Read setup.json (located in .claude/skills/browser-automation/setup.json)
2. Check setupComplete field:

- If true: All prerequisites are met, proceed with browser commands - If false: Setup required - follow the steps below

If Setup is Required (setupComplete: false)

Run these commands in the plugin directory:

# 1. Install dependencies and build (REQUIRED)
# This automatically builds TypeScript
npm install
# or: pnpm install
# or: bun install

# 2. Link the browser command globally (REQUIRED)
npm link

# 3. Configure API key (REQUIRED)
# Option 1 (RECOMMENDED): Export in your terminal
export ANTHROPIC_API_KEY="your-api-key-here"

# Option 2: Or use .env file
cp .env.example .env
# Then edit .env and add: ANTHROPIC_API_KEY="your-api-key-here"

# 4. Verify Chrome is installed
# Chrome should be at standard location for your OS

# 5. Test the installation
browser navigate https://example.com

# 6. If test succeeds, update setup.json
# Set all "installed"/"configured" fields to true
# Set "setupComplete" to true

Prerequisites Summary

• ✅ Google Chrome installed on your system
• ✅ Node.js dependencies installed and TypeScript built (npm install runs build automatically)
• ✅ Browser command globally available (npm link creates the global symlink)
• ✅ Anthropic API key configured (exported as ANTHROPIC_API_KEY environment variable or in .env file)

DO NOT attempt to use browser commands if setupComplete: false in setup.json. Guide the user through setup first.

Available Commands

Navigate to URLs

browser navigate <url>

When to use: Opening any website, loading a specific URL, going to a web page.

Example usage:

• browser navigate https://example.com
• browser navigate https://news.ycombinator.com

Output: JSON with success status, message, and screenshot path

Interact with Pages

browser act "<action>"

When to use: Clicking buttons, filling forms, scrolling, selecting options, typing text.

Example usage:

• browser act "click the Sign In button"
• browser act "fill in the email field with test@example.com"
• browser act "scroll down to the footer"
• browser act "type 'laptop' in the search box and press enter"

Important: Be as specific as possible - details make a world of difference. When filling fields, you don't need to combine 'click and type'; the tool will perform a fill similar to Playwright's fill function.

Output: JSON with success status, message, and screenshot path

Extract Data

browser extract "<instruction>" ['{"field": "type"}']

When to use: Scraping data, getting specific information, collecting structured content.

Schema format (optional): JSON object where keys are field names and values are types:

• "string" for text
• "number" for numeric values
• "boolean" for true/false values

Note: The schema parameter is optional. If omitted or if schema validation fails, extraction will proceed without type validation.

Example usage:

• browser extract "get the product title and price" '{"title": "string", "price": "number"}'
• browser extract "get all article headlines" '{"headlines": "string"}'
• browser extract "get the page title" (no schema)

Output: JSON with success status, extracted data, and screenshot path

Discover Elements

browser observe "<query>"

When to use: Understanding page structure, finding what's clickable, discovering form fields.

Example usage:

• browser observe "find all clickable buttons"
• browser observe "find all form fields"
• browser observe "find all navigation links"

Output: JSON with success status, discovered elements, and screenshot path

Take Screenshots

browser screenshot

When to use: Visual verification, documenting page state, debugging, creating records.

Notes:

• Screenshots are saved to the plugin directory's agent/browser_screenshots/ folder
• Images larger than 2000x2000 pixels are automatically resized
• Filename includes timestamp for uniqueness

Output: JSON with success status and screenshot path

Clean Up

browser close

When to use: After completing all browser interactions, to free up resources.

Output: JSON with success status and message

Browser Behavior

Persistent Browser: The browser stays open between commands for faster sequential operations and to preserve browser state (cookies, sessions, etc.).

Reuse Existing: If Chrome is already running on port 9222, it will reuse that instance.

Minimized Launch: Chrome opens off-screen (position -9999,-9999) to avoid disrupting workflow.

Safe Cleanup: The browser only closes when you explicitly call the close command.

Best Practices

1. Always navigate first: Before interacting with a page, navigate to the URL
2. 📸 Always view screenshots: After each command (navigate, act, extract, observe), use the Read tool to view the screenshot and verify the command worked correctly
3. Use natural language: Describe actions as you would instruct a human
4. Extract with clear schemas: Define field names and types explicitly in JSON
5. Handle errors gracefully: Check the success field in JSON output; if an action fails, view the screenshot and try using observe to understand the page better
6. Close when done: Always clean up browser resources after completing tasks
7. Be specific: Use precise selectors in natural language ("the blue Submit button" vs "the button")
8. Chain commands: Run multiple commands sequentially without reopening the browser

Common Patterns

Simple browsing task

browser navigate https://example.com
browser act "click the login button"
browser screenshot
browser close

Data extraction task

browser navigate https://example.com/products
browser act "wait for page to load"
browser extract "get all products" '{"name": "string", "price": "number"}'
# Or without schema:
# browser extract "get the page content"
browser close

Multi-step interaction

browser navigate https://example.com/login
browser act "fill in email with user@example.com"
browser act "fill in password with mypassword"
browser act "click the submit button"
browser screenshot
browser close

Debugging workflow

browser navigate https://example.com
browser screenshot
browser observe "find all buttons"
browser act "click the specific button"
browser screenshot
browser close

Troubleshooting

Page not loading: Wait a few seconds after navigation before acting. You can explicitly: browser act "wait for the page to fully load"

Element not found: Use observe to discover what elements are actually available on the page

Action fails: Be more specific in natural language description. Instead of "click the button", try "click the blue Submit button in the form"

Screenshots missing: Check the plugin directory's agent/browser_screenshots/ folder for saved files

Chrome not found: Install Google Chrome or the CLI will show an error with installation instructions

Port 9222 in use: Another Chrome debugging session is running. Close it or wait for timeout

For detailed examples, see EXAMPLES.md. For API reference and technical details, see REFERENCE.md.

Dependencies

To use this skill, install these dependencies only if they aren't already present:

npm install
# or
pnpm install
# or
bun install