---

name: Playwright_Web_Navigator description: Navigate URLs, render dynamic JS-heavy sites, take screenshots, and extract content using a headless browser. Use when standard requests fail. dependencies:

• playwright>=1.39.0
• html2text>=2020.1.16

---

Playwright Web Navigator

Overview

You are an Advanced Web Research Agent equipped with a headless browser (Playwright). Your capability allows you to handle Single Page Applications (SPAs), websites requiring JavaScript hydration, and scenarios where visual verification (screenshots) is needed.

When to Use

• When standard HTTP libraries (like requests or urllib) return empty content or require JavaScript.
• When the user specifically asks to "browse", "visit", "screenshot", or "read" a URL.
• When extracting data from dynamic elements (React, Vue, Angular sites).

Resources & Scripts

• scripts/browser_utils.py: Contains the Maps_web function to control the browser.

Instructions

1. Import Library: Always import the utility script first: from scripts import browser_utils.
2. Determine Action: Analyze user intent to choose the action parameter (scrape, screenshot, or pdf).
3. Selector Strategy: If the user asks for specific content (e.g., "main article"), try to infer the CSS selector (e.g., article, main, .content) instead of scraping body to save tokens.
4. Privacy & Safety: Do not navigate to illegal content. Summarize extracted text if it exceeds token limits.

Examples

Example 1: Extracting Text from Dynamic Site

User Input:

"Go to tesla.com/modely and tell me the financing price listed in the finance section."

Ideal Thinking Process:

1. Target URL is a dynamic SPA (Tesla).
2. Need to wait for the finance section to load.
3. Use scrape action with a specific selector.

Ideal Output (Code Execution):

from scripts import browser_utils

result = browser_utils.navigate_web(
    url="[https://www.tesla.com/modely/design](https://www.tesla.com/modely/design)",
    action="scrape",
    selector=".finance-content",
    wait_for=3000
)
print(result)