full-page-screenshot
About
This skill captures full-page screenshots of web pages as PNG images, handling both open browser tabs and specified URLs. It automatically manages SPA scroll containers, lazy-loaded images, and stitches together long pages. The tool connects directly to Chrome DevTools Protocol without external dependencies.
Quick Install
Claude Code
Recommendednpx skills add LewisLiu007/full-page-screenshot -a claude-code/plugin add https://github.com/LewisLiu007/full-page-screenshotgit clone https://github.com/LewisLiu007/full-page-screenshot.git ~/.claude/skills/full-page-screenshotCopy and paste this command in Claude Code to install this skill
Documentation
Full Page Screenshot
对浏览器中的网页生成一张完整的长截图 PNG,覆盖页面全部内容(包括需要滚动才能看到的部分)。直连 Chrome DevTools WebSocket,无需 CDP Proxy 或其他中间层。
前置条件
只需 Node.js 22+ 和 Chrome 开启远程调试:
node "${CLAUDE_SKILL_DIR}/scripts/full-page-screenshot.mjs" --check
如果 chrome 检查失败,引导用户:打开 chrome://inspect/#remote-debugging,勾选 "Allow remote debugging for this browser instance"。
工作流程
用户描述了已打开的 tab(推荐,尤其是需要登录的页面)
先列出所有 tab,匹配用户描述找到 targetId:
node "${CLAUDE_SKILL_DIR}/scripts/full-page-screenshot.mjs" --list
然后截图该 tab:
node "${CLAUDE_SKILL_DIR}/scripts/full-page-screenshot.mjs" <targetId> <output> [--width N] [--dpr N]
用户给了 URL
一条命令完成全部流程(建 tab → 等加载 → 截图 → 关 tab):
node "${CLAUDE_SKILL_DIR}/scripts/full-page-screenshot.mjs" --url "<URL>" <output> [--width N] [--dpr N] [--wait N]
注意:--url 模式创建后台 tab。对于需要登录的内部系统(如 SAP、企业 SSO),后台 tab 可能无法完成认证。此时应引导用户先在 Chrome 中打开页面,然后使用 targetId 模式截图。
参数说明
| 参数 | 说明 | 默认值 |
|---|---|---|
output | 输出文件路径 | /tmp/screenshot.png |
--width | 视口宽度(CSS 像素)。文章页 1200,宽表格/面板 1440-1920 | 1200 |
--dpr | 设备像素比。2 可获 Retina 清晰度,但体积翻 4 倍 | 1 |
--wait | 页面加载超时(毫秒),仅 --url 模式 | 15000 |
验证输出
sips -g pixelWidth -g pixelHeight <output>
裁剪局部验证(需要 Pillow):
from PIL import Image
img = Image.open("<output>")
w, h = img.size
img.crop((0, 0, w, 1500)).save("/tmp/check_top.png")
img.crop((0, h-1500, w, h)).save("/tmp/check_bottom.png")
核心能力
- SPA 内部滚动容器:自动检测
overflow-y: auto/scroll的容器,用!important覆盖 CSS 高度约束(含 Tailwindh-[calc(...)]等),展开全部内容后截图 - DOM 稳定等待:
readyState=complete后继续监测 DOM 元素数量变化,直到 SPA 数据渲染完成 - 懒加载处理:逐步滚动触发 IntersectionObserver,等待所有
img.complete - 超长页面分片:>16000px 的页面自动分片截取(每片 8000px),用 Python PIL 拼接
- Proxy 兼容:当 web-access CDP Proxy 运行时,自动检测并通过 proxy API 完成截图(滚动容器分片拼接)
注意事项
- 超长页面(> 20000px)在高 DPR 下可能导致 Chrome 内存问题。遇到此情况降低
--dpr到 1 或减小--width。 --url模式会自动创建后台 tab 并在截图后关闭,不影响用户已有 tab。- 截图期间会临时改变页面视口尺寸和滚动容器样式,完成后自动恢复。
- 脚本通过读取
DevToolsActivePort文件发现 Chrome 调试端口,直连 WebSocket,无需任何代理或中间服务。 - 分片拼接需要 Python 3 + Pillow(macOS 通常已安装)。无 Pillow 时会保留分片文件。
- 当 web-access 的 CDP Proxy 在运行时,脚本先尝试直连,超时后降级到 proxy API(通过 /eval 和 /screenshot 端点操作)。
GitHub Repository
Related Skills
llamaguard
OtherLlamaGuard is Meta's 7-8B parameter model for moderating LLM inputs and outputs across six safety categories like violence and hate speech. It offers 94-95% accuracy and can be deployed using vLLM, Hugging Face, or Amazon SageMaker. Use this skill to easily integrate content filtering and safety guardrails into your AI applications.
cost-optimization
OtherThis Claude Skill helps developers optimize cloud costs through resource rightsizing, tagging strategies, and spending analysis. It provides a framework for reducing cloud expenses and implementing cost governance across AWS, Azure, and GCP. Use it when you need to analyze infrastructure costs, right-size resources, or meet budget constraints.
quantizing-models-bitsandbytes
OtherThis skill quantizes LLMs to 8-bit or 4-bit precision using bitsandbytes, achieving 50-75% memory reduction with minimal accuracy loss. It's ideal for running larger models on limited GPU memory or accelerating inference, supporting formats like INT8, NF4, and FP4. The skill integrates with HuggingFace Transformers and enables QLoRA training and 8-bit optimizers.
dispatching-parallel-agents
OtherThis Claude Skill dispatches multiple agents to investigate and fix 3+ independent problems concurrently. It is designed for scenarios involving unrelated failures that can be resolved without shared state or dependencies. The core capability is parallel problem-solving, assigning one agent per independent problem domain to maximize efficiency.
