重要前提
安装AI Skills的关键前提是:必须科学上网,且开启TUN模式,这一点至关重要,直接决定安装能否顺利完成,在此郑重提醒三遍:科学上网,科学上网,科学上网。查看完整安装教程 →
geo-technical by zubair-trabzada/geo-seo-claude
npx skills add https://github.com/zubair-trabzada/geo-seo-claude --skill geo-technical技术 SEO 构成了传统搜索可见性和 AI 搜索引用的基础。一个技术上有问题的网站无法被任何平台抓取、索引或引用。本技能审计 8 个类别的技术健康状况,特别关注 GEO 要求——最关键的是服务端渲染(AI 爬虫不执行 JavaScript)和AI 爬虫访问(许多网站无意中在 robots.txt 中屏蔽了 AI 爬虫)。
https://[domain]/robots.txtUser-agent、Allow、Disallow 指令Sitemap: https://[domain]/sitemap.xml广告位招租
在这里展示您的产品或服务
触达数万 AI 开发者,精准高效
检查 robots.txt 中针对以下 AI 爬虫的指令:
| 爬虫 | User-Agent | 平台 |
|---|---|---|
| GPTBot | GPTBot | ChatGPT / OpenAI |
| Google-Extended | Google-Extended | Gemini / Google AI 训练 |
| Googlebot | Googlebot | Google 搜索 + AI 概览 |
| Bingbot | bingbot | Bing Copilot + ChatGPT(通过 Bing) |
| PerplexityBot | PerplexityBot | Perplexity AI |
| ClaudeBot | ClaudeBot | Anthropic Claude |
| Amazonbot | Amazonbot | Alexa / Amazon AI |
| CCBot | CCBot | Common Crawl(被许多 AI 模型使用) |
| FacebookBot | FacebookExternalHit | Meta AI |
| Bytespider | Bytespider | TikTok / ByteDance AI |
| Applebot-Extended | Applebot-Extended | Apple Intelligence |
AI 爬虫访问评分:
重要细节:屏蔽 Google-Extended 并不屏蔽 Googlebot。Google-Extended 仅控制 AI 训练数据的使用,而非搜索索引。然而,屏蔽 Google-Extended 可能会减少在 AI 概览中的出现。除非有特定的数据许可担忧,否则建议允许 Google-Extended。
/sitemap.xml、/sitemap_index.xml)<lastmod> 日期(应存在且准确)<meta name="robots" content="noindex">X-Robots-Tag: noindex HTTP 头类别评分:
| 检查项 | 分数 |
|---|---|
| robots.txt 有效且完整 | 3 |
| 允许 AI 爬虫 | 5 |
| XML 站点地图存在且有效 | 3 |
| 抓取深度在 3 次点击内 | 2 |
| 没有错误的 noindex 指令 | 2 |
<link rel="canonical" href="..."> 标签?sort=price 创建重复页面)rel="next" / rel="prev"(注意:自 2019 年起 Google 忽略这些,但 Bing 仍在使用)rel="canonical" 指向一个“查看全部”页面或第一页<link rel="alternate" hreflang="xx"> 标签site:domain.com 估算)类别评分:
| 检查项 | 分数 |
|---|---|
| 所有页面上的规范标签正确 | 3 |
| 无重复内容问题 | 3 |
| 分页处理正确 | 2 |
| Hreflang 正确(如适用) | 2 |
| 无索引膨胀 | 2 |
检查 HTTP 响应头中是否有:
| 头信息 | 要求值 | 目的 |
|---|---|---|
Strict-Transport-Security | max-age=31536000; includeSubDomains | 强制使用 HTTPS |
Content-Security-Policy | 适当的策略 | 防止 XSS |
X-Content-Type-Options | nosniff | 防止 MIME 嗅探 |
X-Frame-Options | DENY 或 SAMEORIGIN | 防止点击劫持 |
Referrer-Policy | strict-origin-when-cross-origin 或更严格 | 控制引荐来源数据 |
Permissions-Policy | 适当的限制 | 控制浏览器功能 |
类别评分:
| 检查项 | 分数 |
|---|---|
| 使用有效证书强制实施 HTTPS | 4 |
| 存在 HSTS 头 | 2 |
| X-Content-Type-Options | 1 |
| X-Frame-Options | 1 |
| Referrer-Policy | 1 |
| Content-Security-Policy | 1 |
/blog/seo-guide 而不是 /blog?id=12345/category/subcategory/pagerobots.txt 的 Disallow 指令类别评分:
| 检查项 | 分数 |
|---|---|
| 简洁、可读的 URL | 2 |
| 逻辑层次结构 | 2 |
| 无重定向链(最多 1 次跳转) | 2 |
| 已配置参数处理 | 2 |
自 2024 年 7 月 起,Google 仅使用移动版 Googlebot 抓取所有网站。不再有桌面版抓取。如果你的网站在移动端无法工作,那么它对 Google 就无法工作。就是这样。
<meta name="viewport" content="width=device-width, initial-scale=1">类别评分:
| 检查项 | 分数 |
|---|---|
| viewport meta 标签正确 | 3 |
| 响应式布局(无水平滚动) | 3 |
| 点击目标大小适当 | 2 |
| 字体大小清晰易读 | 2 |
核心 Web 指标使用真实用户数据(现场数据)的第 75 百分位数作为基准。实验室数据对调试有用,但现场数据决定了排名信号。
| 指标 | 良好 | 需要改进 | 较差 | 说明 |
|---|---|---|---|---|
| LCP (最大内容绘制) | < 2.5s | 2.5s - 4.0s | > 4.0s | 衡量加载——直到最大可见元素渲染的时间 |
| INP (交互到下次绘制) | < 200ms | 200ms - 500ms | > 500ms | 于 2024 年 3 月取代 FID。衡量所有交互,而不仅仅是首次 |
| CLS (累积布局偏移) | < 0.1 | 0.1 - 0.25 | > 0.25 | 衡量视觉稳定性——意外的布局移动 |
当无法获取真实用户数据时,根据页面特征进行估算:
<link rel="preload"> 预加载requestIdleCallback 或 scheduler.yield() 将长任务(>50ms)分解为更小的块content-visibility: autowidth 和 height 属性aspect-ratio 或显式尺寸为广告和嵌入内容预留空间font-display: swap 并配合调整大小的后备字体类别评分:
| 检查项 | 分数 |
|---|---|
| LCP < 2.5s | 5 |
| INP < 200ms | 5 |
| CLS < 0.1 | 5 |
AI 爬虫(GPTBot、PerplexityBot、ClaudeBot 等)不执行 JavaScript。它们获取原始 HTML 并解析它。如果你的内容是由 React、Vue、Angular 或任何其他 JavaScript 框架在客户端渲染的,AI 爬虫看到的将是一个空页面。
即使是执行 JavaScript 的 Googlebot,也会因为需要额外的抓取预算而降低 JS 渲染内容的优先级。Google 在一个单独的“渲染队列”中处理 JS 渲染,这可能会使索引延迟数天或数周。
curl -s [URL]| 框架 | SSR 解决方案 |
|---|---|
| React | Next.js (SSR/SSG), Remix, Gatsby (SSG) |
| Vue | Nuxt.js (SSR/SSG) |
| Angular | Angular Universal |
| Svelte | SvelteKit |
| 通用 | Prerender.io (预渲染服务), Rendertron |
类别评分:
| 检查项 | 分数 |
|---|---|
| 主要内容在原始 HTML 中 | 8 |
| Meta 标签 + 结构化数据在原始 HTML 中 | 4 |
| 内部链接在原始 HTML 中 | 3 |
curl -o /dev/null -s -w 'TTFB: %{time_starttransfer}s\n' [URL]loading="lazy"async 或 defer)<head> 中的渲染阻塞资源Cache-Control 头max-age=31536000(1 年)并配合内容哈希的文件名no-cache(ETag 或 Last-Modified)CF-Ray (Cloudflare)、X-Cache (AWS CloudFront)、X-Served-By (Fastly)类别评分:
| 检查项 | 分数 |
|---|---|
| TTFB < 800ms | 3 |
| 页面大小 < 2MB | 2 |
| 图片优化(格式、大小、懒加载) | 3 |
| JS 包大小合理(< 200KB 压缩) | 2 |
| 启用压缩(gzip/brotli) | 2 |
| 静态资源上的缓存头 | 2 |
| 使用 CDN | 1 |
IndexNow 是一个开放协议,允许网站在内容创建、更新或删除时立即通知搜索引擎。由 Bing、Yandex、Seznam 和 Naver 支持。Google 不支持 IndexNow,但会监控该协议。
ChatGPT 使用 Bing 的索引。Bing Copilot 使用 Bing 的索引。更快的 Bing 索引意味着在两大主要平台上更快的 AI 可见性。
https://[domain]/.well-known/indexnow-key.txt 或类似文件| 类别 | 最高分 | 权重 |
|---|---|---|
| 可抓取性 | 15 | 核心基础 |
| 可索引性 | 12 | 核心基础 |
| 安全性 | 10 | 信任信号 |
| URL 结构 | 8 | 抓取效率 |
| 移动端优化 | 10 | Google 要求 |
| 核心 Web 指标 | 15 | 排名信号 |
| 服务端渲染 | 15 | GEO 关键 |
| 页面速度与服务器 | 15 | 性能 |
| 总计 | 100 |
生成 GEO-TECHNICAL-AUDIT.md,内容如下:
# GEO Technical SEO Audit — [Domain]
Date: [Date]
## Technical Score: XX/100
## Score Breakdown
| Category | Score | Status |
|---|---|---|
| Crawlability | XX/15 | Pass/Warn/Fail |
| Indexability | XX/12 | Pass/Warn/Fail |
| Security | XX/10 | Pass/Warn/Fail |
| URL Structure | XX/8 | Pass/Warn/Fail |
| Mobile Optimization | XX/10 | Pass/Warn/Fail |
| Core Web Vitals | XX/15 | Pass/Warn/Fail |
| Server-Side Rendering | XX/15 | Pass/Warn/Fail |
| Page Speed & Server | XX/15 | Pass/Warn/Fail |
Status: Pass = 80%+ of category points, Warn = 50-79%, Fail = <50%
## AI Crawler Access
| Crawler | User-Agent | Status | Recommendation |
|---|---|---|---|
| GPTBot | GPTBot | Allowed/Blocked | [Action] |
| Googlebot | Googlebot | Allowed/Blocked | [Action] |
[Continue for all AI crawlers]
## Critical Issues (fix immediately)
[List with specific page URLs and what is wrong]
## Warnings (fix this month)
[List with details]
## Recommendations (optimize this quarter)
[List with details]
## Detailed Findings
[Per-category breakdown with evidence]
每周安装次数
56
仓库
GitHub Stars
3.8K
首次出现
Feb 27, 2026
安全审计
安装于
codex55
opencode55
kimi-cli53
gemini-cli53
amp53
cline53
Technical SEO forms the foundation of both traditional search visibility and AI search citation. A technically broken site cannot be crawled, indexed, or cited by any platform. This skill audits 8 categories of technical health with specific attention to GEO requirements — most critically, server-side rendering (AI crawlers do not execute JavaScript) and AI crawler access (many sites inadvertently block AI crawlers in robots.txt).
https://[domain]/robots.txtUser-agent, Allow, Disallow directivesSitemap: https://[domain]/sitemap.xmlCheck robots.txt for directives targeting these AI crawlers:
| Crawler | User-Agent | Platform |
|---|---|---|
| GPTBot | GPTBot | ChatGPT / OpenAI |
| Google-Extended | Google-Extended | Gemini / Google AI training |
| Googlebot | Googlebot | Google Search + AI Overviews |
| Bingbot | bingbot | Bing Copilot + ChatGPT (via Bing) |
| PerplexityBot | PerplexityBot | Perplexity AI |
| ClaudeBot | ClaudeBot | Anthropic Claude |
| Amazonbot | Amazonbot | Alexa / Amazon AI |
| CCBot | CCBot | Common Crawl (used by many AI models) |
| FacebookBot | FacebookExternalHit | Meta AI |
| Bytespider |
Scoring for AI crawler access:
Important nuance : Blocking Google-Extended does NOT block Googlebot. Google-Extended only controls AI training data usage, not search indexing. However, blocking Google-Extended may reduce presence in AI Overviews. Recommend allowing Google-Extended unless there is a specific data licensing concern.
/sitemap.xml, /sitemap_index.xml)<lastmod> dates (should be present and accurate)<meta name="robots" content="noindex"> on pages that SHOULD be indexedX-Robots-Tag: noindex HTTP headersCategory Scoring:
| Check | Points |
|---|---|
| robots.txt valid and complete | 3 |
| AI crawlers allowed | 5 |
| XML sitemap present and valid | 3 |
| Crawl depth within 3 clicks | 2 |
| No erroneous noindex directives | 2 |
<link rel="canonical" href="..."> tag?sort=price creating duplicate pages)rel="next" / rel="prev" (note: Google ignores these as of 2019, but Bing still uses them)rel="canonical" on paginated pages pointing to a view-all page or the first page<link rel="alternate" hreflang="xx"> tagssite:domain.com estimate)Category Scoring:
| Check | Points |
|---|---|
| Canonical tags correct on all pages | 3 |
| No duplicate content issues | 3 |
| Pagination handled correctly | 2 |
| Hreflang correct (if applicable) | 2 |
| No index bloat | 2 |
Check HTTP response headers for:
| Header | Required Value | Purpose |
|---|---|---|
Strict-Transport-Security | max-age=31536000; includeSubDomains | Forces HTTPS |
Content-Security-Policy | Appropriate policy | Prevents XSS |
X-Content-Type-Options | nosniff | Prevents MIME sniffing |
X-Frame-Options | or |
Category Scoring:
| Check | Points |
|---|---|
| HTTPS enforced with valid cert | 4 |
| HSTS header present | 2 |
| X-Content-Type-Options | 1 |
| X-Frame-Options | 1 |
| Referrer-Policy | 1 |
| Content-Security-Policy | 1 |
/blog/seo-guide not /blog?id=12345/category/subcategory/pagerobots.txt Disallow for parameter variationsCategory Scoring:
| Check | Points |
|---|---|
| Clean, readable URLs | 2 |
| Logical hierarchy | 2 |
| No redirect chains (max 1 hop) | 2 |
| Parameter handling configured | 2 |
As of July 2024 , Google crawls ALL sites exclusively with mobile Googlebot. There is no desktop crawling. If your site does not work on mobile, it does not work for Google. Period.
<meta name="viewport" content="width=device-width, initial-scale=1">Category Scoring:
| Check | Points |
|---|---|
| Viewport meta tag correct | 3 |
| Responsive layout (no horizontal scroll) | 3 |
| Tap targets appropriately sized | 2 |
| Font sizes legible | 2 |
Core Web Vitals use the 75th percentile of real user data (field data) as the benchmark. Lab data is useful for debugging but field data determines the ranking signal.
| Metric | Good | Needs Improvement | Poor | Notes |
|---|---|---|---|---|
| LCP (Largest Contentful Paint) | < 2.5s | 2.5s - 4.0s | > 4.0s | Measures loading — time until largest visible element renders |
| INP (Interaction to Next Paint) | < 200ms | 200ms - 500ms | > 500ms | Replaced FID in March 2024. Measures ALL interactions, not just first |
| CLS (Cumulative Layout Shift) | < 0.1 | 0.1 - 0.25 | > 0.25 | Measures visual stability — unexpected layout movements |
When real user data is unavailable, estimate from page characteristics:
<link rel="preload">requestIdleCallback or scheduler.yield()content-visibility: auto for off-screen contentwidth and height attributes on images and videosaspect-ratio or explicit dimensionsfont-display: swap with size-adjusted fallback fontsCategory Scoring:
| Check | Points |
|---|---|
| LCP < 2.5s | 5 |
| INP < 200ms | 5 |
| CLS < 0.1 | 5 |
AI crawlers (GPTBot, PerplexityBot, ClaudeBot, etc.) do NOT execute JavaScript. They fetch the raw HTML and parse it. If your content is rendered client-side by React, Vue, Angular, or any other JavaScript framework, AI crawlers see an empty page.
Even Googlebot, which does execute JavaScript, deprioritizes JS-rendered content due to the additional crawl budget required. Google processes JS rendering in a separate "rendering queue" that can delay indexing by days or weeks.
curl -s [URL]| Framework | SSR Solution |
|---|---|
| React | Next.js (SSR/SSG), Remix, Gatsby (SSG) |
| Vue | Nuxt.js (SSR/SSG) |
| Angular | Angular Universal |
| Svelte | SvelteKit |
| Generic | Prerender.io (prerendering service), Rendertron |
Category Scoring:
| Check | Points |
|---|---|
| Main content in raw HTML | 8 |
| Meta tags + structured data in raw HTML | 4 |
| Internal links in raw HTML | 3 |
curl -o /dev/null -s -w 'TTFB: %{time_starttransfer}s\n' [URL]loading="lazy"async or defer)<head>Cache-Control headers on static resources (images, CSS, JS)max-age=31536000 (1 year) with content-hashed filenamesno-cache with validation (ETag or Last-Modified)CF-Ray (Cloudflare), X-Cache (AWS CloudFront), X-Served-By (Fastly)Category Scoring:
| Check | Points |
|---|---|
| TTFB < 800ms | 3 |
| Page weight < 2MB | 2 |
| Images optimized (format, size, lazy) | 3 |
| JS bundles reasonable (< 200KB compressed) | 2 |
| Compression enabled (gzip/brotli) | 2 |
| Cache headers on static resources | 2 |
| CDN in use | 1 |
IndexNow is an open protocol that allows websites to notify search engines instantly when content is created, updated, or deleted. Supported by Bing, Yandex, Seznam, and Naver. Google does NOT support IndexNow but monitors the protocol.
ChatGPT uses Bing's index. Bing Copilot uses Bing's index. Faster Bing indexing means faster AI visibility on two major platforms.
https://[domain]/.well-known/indexnow-key.txt or similar| Category | Max Points | Weight |
|---|---|---|
| Crawlability | 15 | Core foundation |
| Indexability | 12 | Core foundation |
| Security | 10 | Trust signal |
| URL Structure | 8 | Crawl efficiency |
| Mobile Optimization | 10 | Google requirement |
| Core Web Vitals | 15 | Ranking signal |
| Server-Side Rendering | 15 | GEO critical |
| Page Speed & Server | 15 | Performance |
| Total | 100 |
Generate GEO-TECHNICAL-AUDIT.md with:
# GEO Technical SEO Audit — [Domain]
Date: [Date]
## Technical Score: XX/100
## Score Breakdown
| Category | Score | Status |
|---|---|---|
| Crawlability | XX/15 | Pass/Warn/Fail |
| Indexability | XX/12 | Pass/Warn/Fail |
| Security | XX/10 | Pass/Warn/Fail |
| URL Structure | XX/8 | Pass/Warn/Fail |
| Mobile Optimization | XX/10 | Pass/Warn/Fail |
| Core Web Vitals | XX/15 | Pass/Warn/Fail |
| Server-Side Rendering | XX/15 | Pass/Warn/Fail |
| Page Speed & Server | XX/15 | Pass/Warn/Fail |
Status: Pass = 80%+ of category points, Warn = 50-79%, Fail = <50%
## AI Crawler Access
| Crawler | User-Agent | Status | Recommendation |
|---|---|---|---|
| GPTBot | GPTBot | Allowed/Blocked | [Action] |
| Googlebot | Googlebot | Allowed/Blocked | [Action] |
[Continue for all AI crawlers]
## Critical Issues (fix immediately)
[List with specific page URLs and what is wrong]
## Warnings (fix this month)
[List with details]
## Recommendations (optimize this quarter)
[List with details]
## Detailed Findings
[Per-category breakdown with evidence]
Weekly Installs
56
Repository
GitHub Stars
3.8K
First Seen
Feb 27, 2026
Security Audits
Gen Agent Trust HubPassSocketPassSnykWarn
Installed on
codex55
opencode55
kimi-cli53
gemini-cli53
amp53
cline53
程序化SEO实战指南:大规模创建优质页面,避免内容单薄惩罚
42,700 周安装
| Bytespider |
| TikTok / ByteDance AI |
| Applebot-Extended | Applebot-Extended | Apple Intelligence |
DENYSAMEORIGIN| Prevents clickjacking |
Referrer-Policy | strict-origin-when-cross-origin or stricter | Controls referrer data |
Permissions-Policy | Appropriate restrictions | Controls browser features |