Web Fetch Skill

Fetch web content using the best available method, in priority order:

1. WebFetch Tool (Primary — always try this first)

Use the built-in WebFetch tool for all web page fetching. It renders pages in a real Electron BrowserWindow with full JavaScript execution, handles SPAs, dynamic content, anti-bot protections, and returns clean markdown. This is NOT a simple HTTP request.

WebFetch(url="https://example.com/article", prompt="Extract the main content")

• Handles JavaScript-rendered pages (React, Vue, etc.)
• Shares session cookies (can access logged-in content)
• Auto-extracts article content via Readability
• Returns clean markdown, ready to use

Always start with WebFetch. Only move to the next option if WebFetch fails or is insufficient.

2. Browser Skill (Fallback for interactive pages)

If WebFetch fails (e.g., page requires login interaction, CAPTCHA, or multi-step navigation), use the alma browser skill:

# Open URL in a real Chrome browser tab
alma browser open "https://example.com"

# Read the page content as markdown
alma browser read <tabId>

# For pages that need interaction first
alma browser click <tabId> "button.load-more"
alma browser read <tabId>

Use the browser skill when you need:

• To interact with the page (click, scroll, fill forms) before reading
• Access to the user's authenticated Chrome sessions
• To handle CAPTCHAs or complex anti-bot pages

3. curl (Last resort)

Only use curl when both WebFetch and browser skill are unavailable or for simple API/JSON endpoints:

# JSON API (curl is fine for APIs)
curl -s "https://api.example.com/data" | jq '.'

# API with auth headers
curl -s -H "Authorization: Bearer TOKEN" -H "Accept: application/json" "URL"

# Download a file
curl -sL -o /tmp/file.pdf "URL"

# Check HTTP status
curl -sL -o /dev/null -w "%{http_code}" "URL"

Do NOT use curl for regular web pages — it cannot execute JavaScript and gets blocked by most modern sites. Use WebFetch instead.

Tips

• WebFetch first, always. It handles 90%+ of web fetching needs.
• For JSON APIs, curl is acceptable since no JS rendering is needed.
• Pipe curl output to head -N to limit output size.
• Use jq for JSON responses.