web-fetch

name	web-fetch
description	Fetch and read web pages, APIs, and online content. Use when users share URLs or ask about web content.
allowed-tools	["Bash","WebFetch","ChromeRelayStatus","ChromeRelayNavigate","ChromeRelayRead","ChromeRelayReadDom","ChromeRelayListTabs","ChromeRelayClick","ChromeRelayScreenshot","ChromeRelayScroll","ChromeRelayEval"]

Web Fetch Skill

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

1. Relay Check

Before fetching, check whether Chrome Relay is healthy and attached:

ChromeRelayStatus()

• If relay is healthy, continue with the Chrome Relay path below
• If relay is not healthy, skip straight to WebFetch and read the page by URL

2. Chrome Relay (Primary — always try this first)

Use Chrome Relay tools for all web fetching when it is connected. Chrome Relay controls the user's real Chrome browser with existing sessions, cookies, and logins — critical for sites that require authentication (Twitter/X, GitHub, email, etc.).

If relay is healthy, reuse the existing tabs:

ChromeRelayListTabs()

# List open tabs
ChromeRelayListTabs()

# Navigate to a URL (opens in existing or new tab)
ChromeRelayNavigate(url="https://x.com/notifications", tabId=<id>)

# Read page content as clean text
ChromeRelayRead(tabId=<id>)

# Read page DOM structure
ChromeRelayReadDom(tabId=<id>)

# Take a screenshot
ChromeRelayScreenshot(tabId=<id>)

# Click elements
ChromeRelayClick(tabId=<id>, selector="button.load-more")

# Scroll the page
ChromeRelayScroll(tabId=<id>, direction="down")

Always start with Chrome Relay. It handles:

• Authenticated pages (Twitter/X, GitHub, Google, etc.)
• JavaScript-rendered SPAs
• Anti-bot protections (using real Chrome fingerprint)
• Pages behind login walls

3. WebFetch Tool (Fallback for simple/public pages)

If Chrome Relay is not connected or the page is simple public content, fall back to WebFetch:

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

• Renders pages in Electron BrowserWindow with JavaScript
• Good for public articles, docs, APIs
• No authentication/login capability

Tips

• Chrome Relay first, always. It has the user's real login sessions.
• If relay is not connected or healthy, skip straight to WebFetch and read the page by URL.
• WebFetch only for public pages when Chrome Relay is unavailable.
• For Twitter/X: ALWAYS use Chrome Relay — WebFetch cannot access authenticated Twitter content.
• Check ChromeRelayListTabs() first to see if Chrome is connected and find existing tabs.