菜单
Automate Chrome for web scraping, SEO analysis, and extracting data from JavaScript-heavy sites. Use for multi-page scraping, authenticated content, screenshots, or when WebFetch isn't enough.
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。
基于 SOC 职业分类
Manage Gmail - send, read, search emails, manage labels and drafts. Use when user wants to interact with their Gmail account for email operations.
Analyzes UXR report PDFs and generates structured analysis documents with key findings, recommendations, and strategic connections. Use when given a UXR report to analyze.
Render a video clip with captions overlaid, using the Remotion captioner at `/Users/jvincent/Projects/remotion-captioner/`. Use when user provides a video file and wants to add captions/subtitles, mentions "caption this video", "add captions", "burn in subtitles", or provides a video + SRT file pair.
Create and iterate on MARP Markdown slide decks with HubSpot styling (two themes available — Trellis product palette and HubSpot brand-identity palette), AI-generated images via LiteLLM/Gemini, and slidegauge linting. Use when the user wants to create a presentation, build slides, make a slide deck, or work on a talk.
Generate a self-improving "autoresearch" loop for any Claude skill. Given a skill to improve and 3-5 quality constraints, scaffolds a standalone project (modeled on karpathy/autoresearch) that iterates on the skill's prompt by generating N samples per cycle, evaluating them against the constraints, and editing the skill to fix recurring failures — up to 30 cycles or until results are consistently perfect. Use when the user says something like "make an autoresearcher for my X skill" or "build a self-improving loop for skill Y".
Manage Google Calendar - search, create, update events and answer calendar questions. Use when user wants to interact with their Google Calendar for scheduling and calendar operations.
| name | Browser Automation |
| description | Automate Chrome for web scraping, SEO analysis, and extracting data from JavaScript-heavy sites. Use for multi-page scraping, authenticated content, screenshots, or when WebFetch isn't enough. |
| version | 1.0.0 |
| dependencies | node>=18.0.0, puppeteer-core^21.0.0 |
| allowed-tools | ["Bash","Read","Write"] |
Lightweight browser automation using 6 simple Node.js scripts. Use this for web scraping, SEO analysis, or extracting data from JavaScript-heavy sites.
When to use: Multi-page scraping, authenticated content, screenshots, JavaScript-rendered content When to use WebFetch: Simple HTML, single pages, no auth needed
Located in ~/.claude/skills/browser-automation/resources/:
browser-start.js - Launch Chrome with debuggingbrowser-navigate.js <url> - Navigate to URLbrowser-eval.js '<javascript>' - Run JavaScript in page contextbrowser-screenshot.js - Capture screenshotbrowser-cookies.js - Export cookiesbrowser-close.js - Shutdown browsercd ~/.claude/skills/browser-automation/resources
node browser-start.js --headless
node browser-navigate.js "https://example.com"
node browser-eval.js 'Array.from(document.querySelectorAll("h2")).map(h => h.textContent.trim())' --json > headings.json
node browser-close.js
See ~/.claude/skills/browser-automation/README.md for examples and patterns.