---

name: text-browser description: Use text-based browser tools for web browsing, scraping, and automation tasks without GUI. Use when you need to navigate websites, extract content, search the web, or interact with web pages programmatically.

Text-Based Browser

This skill provides guidance for using text-based browser tools to interact with the web without a graphical interface. Perfect for automated scraping, content extraction, web research, and programmatic web interactions.

Available Tools

Web Search

Tool: web_search Use when: You need to search the web for information Parameters:

• query: Search terms
• count: Number of results (1-10)
• country: 2-letter country code for regional results
• search_lang: Language code for results
• freshness: Filter by discovery time

Example:

web_search(query="python web scraping", count=5)

Web Fetch

Tool: web_fetch Use when: You need to extract readable content from a specific URL Parameters:

• url: HTTP or HTTPS URL to fetch
• extractMode: "markdown" or "text" (default: markdown)
• maxChars: Maximum characters to return

Example:

web_fetch(url="https://example.com", extractMode="markdown", maxChars=10000)

Browser Automation

Tool: browser Use when: You need to navigate, click, type, or interact with a website Actions:

• snapshot - Get page structure and elements
• navigate - Go to a URL
• act - Perform actions (click, type, press, select, fill, wait)
• screenshot - Capture page visual
• focus - Switch to a specific tab

Profiles:

• openclaw - Isolated browser (default)
• chrome - Extension relay (requires user to attach tab)

Example:

browser(action="snapshot", profile="openclaw")
browser(action="navigate", targetUrl="https://example.com")
browser(action="act", request={"kind": "click", "ref": "submit-button"})

Workflow Patterns

Pattern 1: Quick Research

Use web_search + web_fetch for fast information gathering:

1. Search for topic
2. Fetch relevant URLs
3. Extract and summarize content

Pattern 2: Page Scraping

Use web_fetch for straightforward content extraction:

1. Get URL
2. Fetch with markdown mode
3. Parse and extract needed information

Pattern 3: Interactive Tasks

Use browser automation for complex interactions:

1. Navigate to page
2. Snapshot to understand structure
3. Perform actions (click, type, wait)
4. Capture results

Pattern 4: Multi-Page Research

Combine tools for comprehensive research:

1. Search for multiple terms
2. Fetch top results
3. Use browser automation for sites requiring JavaScript
4. Aggregate findings

Best Practices

Choose the Right Tool

• web_search - When you don't know specific URLs
• web_fetch - When you have a URL and just need content
• browser - When you need to interact (forms, dynamic content, complex navigation)

Optimize for Speed

• Use web_fetch whenever possible (fastest)
• Prefer snapshot over full-page interactions when possible
• Use refs="aria" for stable element references
• Set appropriate limit and maxChars to avoid bloat

Handle Errors Gracefully

• Network errors: Retry with timeout
• 404/403: Check URL or try alternative
• Parse errors: Try different extract mode or use browser automation
• Rate limiting: Add delays between requests

Respect Rate Limits

• Add delays between repeated requests to same domain
• Use count parameter wisely (3-5 is usually sufficient)
• Cache results when appropriate

Advanced Techniques

Pagination Handling

For paginated content, use browser automation:

1. Snapshot page
2. Extract next button reference
3. Click and wait
4. Repeat until end

Form Submission

Use browser automation for forms:

1. Snapshot to find form fields
2. Use act with kind: "fill" for multiple fields
3. Submit with button click
4. Wait and verify results

Content Extraction Strategy

1. Try web_fetch with markdown mode first (fastest)
2. If insufficient, use browser snapshot for structure
3. Use specific selectors or aria refs for precise extraction
4. Combine with regex parsing if needed

Working with JavaScript Sites

When web_fetch returns incomplete content:

1. Use browser automation
2. Navigate to URL
3. Wait for page load (use act with kind: "wait" and timeMs)
4. Snapshot to get rendered content
5. Extract needed information

Common Use Cases

Product Research

Search for products, compare prices, extract specs:

web_search(query="laptop comparison", count=5)
# Fetch top 3 results
web_fetch(url="result1", extractMode="markdown")
web_fetch(url="result2", extractMode="markdown")
web_fetch(url="result3", extractMode="markdown")

News Aggregation

Gather news from multiple sources:

web_search(query="tech news", count=5, freshness="pd")
web_fetch(url="news1", extractMode="markdown")
web_fetch(url="news2", extractMode="markdown")

Documentation Lookup

Find and extract API docs or guides:

web_search(query="Python requests docs", count=3)
web_fetch(url="docs_url", extractMode="markdown", maxChars=50000)

Data Extraction from Forms

Submit forms and scrape results:

browser(action="navigate", targetUrl="https://example.com/search")
browser(action="act", request={"kind": "fill", "fields": [{"ref": "query", "text": "search term"}]})
browser(action="act", request={"kind": "click", "ref": "search-button"})
browser(action="act", request={"kind": "wait", "timeMs": 2000})
browser(action="snapshot")

Troubleshooting

"Content not loading"

• Switch from web_fetch to browser automation
• Add wait times for dynamic content
• Check for JavaScript rendering requirements

"Can't find element"

• Use refs="aria" for stable references
• Take new snapshot after page changes
• Verify element still exists after actions

"Rate limit errors"

• Add delays between requests (1-3 seconds)
• Reduce count parameter
• Respect robots.txt and terms of service

"Slow performance"

• Use web_fetch instead of browser when possible
• Reduce maxChars to only what's needed
• Cache results when repeating same queries

Security Considerations

• Always respect robots.txt
• Don't overwhelm servers (add delays)
• Don't bypass authentication unless authorized
• Handle sensitive data carefully
• Log actions for transparency

When to Use This Skill

Use text-browser when:

• You need to search the web
• You need to extract content from URLs
• You need to interact with web pages
• You need to scrape or research programmatically
• You need to work with web APIs that don't have direct integration

See Also

• Agent Browser skill - For GUI-based browser automation
• web_search tool - For quick searches
• web_fetch tool - For content extraction