菜单
How to use Chrome DevTools MCP for browser debugging. Use when you need to inspect pages, take screenshots, debug UI issues, or verify visual changes.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/akornmeier/claude-config --skill chrome-debug命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Transform an OpenSpec change phase into UI-ready build prompts through a 3-step pipeline. PRD generation, then UX specification, then build-order prompts. Use when starting implementation of an OpenSpec feature phase, or when the user mentions "openspec-to-prompts", "feature pipeline", or asks to convert a spec phase to implementation prompts. Requires an openspec change-id and phase number as input.
Refine and clarify PRD documentation through structured questioning using the visual AskUserQuestion tool
Convert a rough MVP idea into a demo-grade PRD (Steps 1-7)
基于 SOC 职业分类
正在显示 SKILL.md
| name | chrome-debug |
| description | How to use Chrome DevTools MCP for browser debugging. Use when you need to inspect pages, take screenshots, debug UI issues, or verify visual changes. |
This skill explains how to use the Chrome DevTools MCP for browser debugging and UI verification.
Before using Chrome DevTools MCP, you must launch Chrome in headless mode with remote debugging enabled:
npm run chrome &
This runs Chrome with the required flags for Docker/containerized environments:
--remote-debugging-port=9222 - Enables MCP connection--no-sandbox - Required for Docker--headless - Runs without display--disable-gpu - Avoids GPU issues in containersWait a few seconds for Chrome to start before using MCP tools.
IMPORTANT: Always save screenshots to the .screenshots directory using the filePath parameter:
mcp__chrome-devtools__take_screenshot with filePath: ".screenshots/descriptive-name.png"
Use descriptive filenames like:
.screenshots/homepage-initial.png.screenshots/login-form-error.png.screenshots/after-click-submit.pngThis allows debugging and visual comparison later.
Start Chrome:
npm run chrome &
Wait 2-3 seconds for startup.
Navigate to the dev server:
mcp__chrome-devtools__navigate_page url="http://localhost:5173"
Take a snapshot to see page structure:
mcp__chrome-devtools__take_snapshot
The snapshot shows element UIDs you can use for interactions.
Take a screenshot for visual verification:
mcp__chrome-devtools__take_screenshot filePath=".screenshots/page-state.png"
Interact with elements using UIDs from snapshot:
mcp__chrome-devtools__click uid="1_4"
mcp__chrome-devtools__fill uid="1_5" value="test@example.com"
.screenshots/ directory with descriptive namesnpm run chrome & againwait_for to ensure page content loads before interacting