playwright

pacphi 的头像作者 pacphi

在Display :1上使用Playwright进行浏览器自动化、网页抓取和视觉测试。用于导航网页、点击元素、填写表单、截取屏幕、执行JavaScript和Web应用的视觉验证。通过端口5901上的VNC提供视觉访问

📖 中文导读:在Display :1上使用Playwright进行浏览器自动化、网页抓取和视觉测试。用于导航网页、点击元素、填写表单、截取屏幕、执行JavaScript和Web应用的视觉验证。通过端口5901上的VNC提供视觉访问 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: playwright description: > Browser automation, web scraping, and visual testing with Playwright on Display :1. Use for navigating web pages, clicking elements, filling forms, taking screenshots, executing JavaScript, and visual verification of web applications. Visual access available via VNC on port 5901. version: 2.0.0 author: turbo-flow-claude mcp_server: true protocol: mcp-sdk entry_point: mcp-server/server.js dependencies: - chromium - playwright

Playwright Skill

Browser automation and visual testing via MCP SDK, with direct browser control on Display :1.

When to Use This Skill

  • Navigate to web pages and capture screenshots
  • Click buttons, links, and interactive elements
  • Fill forms and submit data
  • Execute JavaScript in page context
  • Wait for dynamic content and AJAX responses
  • Visual verification of web applications
  • Web scraping and data extraction
  • End-to-end testing of web UIs

Architecture

┌─────────────────────────────┐
│  Claude Code / VisionFlow   │
│  (MCP Client)               │
└──────────────┬──────────────┘
               │ MCP Protocol (stdio)
               ▼
┌─────────────────────────────┐
│  Playwright MCP Server      │
│  (Consolidated from 3 files)│
└──────────────┬──────────────┘
               │ Direct Playwright API
               ▼
┌─────────────────────────────┐
│  Chromium Browser           │
│  Display :1 (VNC 5901)      │
└─────────────────────────────┘

Tools

Tool Description
navigate Navigate browser to URL
screenshot Capture screenshot (full page or viewport)
click Click an element by CSS/XPath selector
type Type text into input field
evaluate Execute JavaScript in page context
wait_for_selector Wait for element to appear/disappear
get_content Get full HTML content of page
get_url Get current page URL and title
close_browser Close browser instance
health_check Check browser connection health

Examples

// Navigate and screenshot
await navigate({ url: "https://example.com" });
await screenshot({ filename: "homepage.png", fullPage: true });

// Fill and submit a form
await type({ selector: "#email", text: "user@example.com" });
await type({ selector: "#password", text: "secret123" });
await click({ selector: "button[type=submit]" });

// Wait for dynamic content
await wait_for_selector({ selector: ".results-loaded" });

// Execute JavaScript
await evaluate({ script: "document.querySelectorAll('.item').length" });

Environment Variables

Variable Default Description
DISPLAY :1 X display for browser
CHROMIUM_PATH /usr/bin/chromium Path to Chromium binary
PLAYWRIGHT_HEADLESS false Run headless (no display)
SCREENSHOT_DIR /tmp/playwright-screenshots Screenshot output directory
PLAYWRIGHT_TIMEOUT 30000 Default timeout in ms
VIEWPORT_WIDTH 1920 Browser viewport width
VIEWPORT_HEIGHT 1080 Browser viewport height

Visual Access

Browser is visible via VNC:

# Connect with VNC client
vncviewer localhost:5901

# Password: turboflow

VisionFlow Integration

This skill exposes playwright://capabilities and playwright://status resources for discovery by VisionFlow's MCP TCP client on port 9500.