` (50–60 chars) and a unique `<meta name="description">` (120–160 chars). Duplicate titles and descriptions cause Google to de-rank or rewrite them. Check both before marking a task done. - Every page MUST render a `<link rel="canonical" href="...">` pointing to its own canonical URL. This is required even on pages that are not duplicated — it is a signal, not just a deduplication tool. - Never generate two pages with the same slug or URL path. Before adding a new content file or route, confirm the path does not already exist in `src/pages/` or the content collection. - Every blog post or article MUST include JSON-LD structured data: at minimum `Article` with `headline`, `datePublished`, `dateModified`, and `author`. Product pages need `Product` schema. FAQ pages need `FAQPage` schema. - Open Graph tags (`og:title`, `og:description`, `og:image`, `og:url`) must be present on every page. The `og:image` must be an absolute URL (not a relative path). - Do NOT use `noindex` on pages that should rank. Do NOT remove `noindex` from pages in `src/pages/api/`, `src/pages/admin/`, or any route that should not be crawled. ## URL and routing conventions - URLs are lowercase, hyphen-separated, no trailing slashes (or consistently with trailing slash if the framework default — pick one and enforce it with a redirect). - Never rename a published URL without adding a 301 redirect from the old path. Broken inbound links are ranking signals that are lost permanently. - Paginated series use `/page/2/` style paths, not query strings (`?page=2`). Query-string pagination is not indexed by Google. ## Content and performance - All images must have descriptive `alt` text that includes the target keyword where natural. Empty `alt=""` is only correct for decorative images. - Images must be served in WebP or AVIF format. No JPEG or PNG without a `<picture>` element providing a next-gen fallback. - Every page must load without render-blocking scripts. No `<script>` without `defer` or `async` in the `<head>` unless it is a critical inline script. - Internal links must use the full path and must not 404. Before adding a link, verify the target page exists. ## Definition of done - `astro check` or `tsc --noEmit` passes. - `astro build` completes without warnings. - Running a spot-check: `curl -s <page-url> | grep -c 'canonical'` returns 1. - No duplicate `<title>` values across built HTML (run `grep -r '<title>'` on dist/). - JSON-LD is present and valid (use schema.org validator). ``` ## 为什么这些规则 - **每页唯一的标题和描述**是内容网站最具影响力的SEO规则。批量生成内容页面的代理通常重复使用相同的元数据模板，生成数十个技术上不同但对爬虫而言看似相同的页面——从而触发软去重惩罚。 - **每个页面都设置规范URL，而不仅仅是重复页面**经常被误解。阅读SEO文档的代理通常只在内容明显重复的地方添加规范URL（例如分页）。实际上，每个页面都应自引用其规范URL，以防止参数注入的爬取变体分散链接权益。 ## 适用场景 - 博客、文档网站、SEO内容中心和营销网站，这些网站以自然搜索为主要获取渠道。 ## 不适用 - 内部工具、仪表盘或应用程序，其中SEO无关紧要—

# AI编码规则针对SEO内容网站 > AGENTS.md规则针对以SEO为重点的内容网站，防止重复元数据，强制执行结构化数据，并防止代理破坏可爬取性。 **Type:** Rule **Tools:** Cursor, Claude Code, Codex, Windsurf **Stack:** Astro, Next.js, TypeScript **Updated:** 2026-06-08 --- 将此文件放在仓库根目录，命名为 `AGENTS.md`。它适用于任何内容密集型网站，这些网站以自然搜索流量为主要增长杠杆——博客、文档网站、营销网站和资源库。 ## AGENTS.md ```md title="AGENTS.md" # Project Rules — SEO Content Site ## Stack - Astro (static) or Next.js (App Router, static export or ISR). - TypeScript strict. Content schema enforced via Zod (content collections or manual). - Tailwind CSS for styling. ## Hard rules — SEO correctness - Every page MUST have a unique `` (50–60 chars) and a unique `<meta name="description">` (120–160 chars). Duplicate titles and descriptions cause Google to de-rank or rewrite them. Check both before marking a task done. - Every page MUST render a `<link rel="canonical" href="...">` pointing to its own canonical URL. This is required even on pages that are not duplicated — it is a signal, not just a deduplication tool. - Never generate two pages with the same slug or URL path. Before adding a new content file or route, confirm the path does not already exist in `src/pages/` or the content collection. - Every blog post or article MUST include JSON-LD structured data: at minimum `Article` with `headline`, `datePublished`, `dateModified`, and `author`. Product pages need `Product` schema. FAQ pages need `FAQPage` schema. - Open Graph tags (`og:title`, `og:description`, `og:image`, `og:url`) must be present on every page. The `og:image` must be an absolute URL (not a relative path). - Do NOT use `noindex` on pages that should rank. Do NOT remove `noindex` from pages in `src/pages/api/`, `src/pages/admin/`, or any route that should not be crawled. ## URL and routing conventions - URLs are lowercase, hyphen-separated, no trailing slashes (or consistently with trailing slash if the framework default — pick one and enforce it with a redirect). - Never rename a published URL without adding a 301 redirect from the old path. Broken inbound links are ranking signals that are lost permanently. - Paginated series use `/page/2/` style paths, not query strings (`?page=2`). Query-string pagination is not indexed by Google. ## Content and performance - All images must have descriptive `alt` text that includes the target keyword where natural. Empty `alt=""` is only correct for decorative images. - Images must be served in WebP or AVIF format. No JPEG or PNG without a `<picture>` element providing a next-gen fallback. - Every page must load without render-blocking scripts. No `<script>` without `defer` or `async` in the `<head>` unless it is a critical inline script. - Internal links must use the full path and must not 404. Before adding a link, verify the target page exists. ## Definition of done - `astro check` or `tsc --noEmit` passes. - `astro build` completes without warnings. - Running a spot-check: `curl -s <page-url> | grep -c 'canonical'` returns 1. - No duplicate `<title>` values across built HTML (run `grep -r '<title>'` on dist/). - JSON-LD is present and valid (use schema.org validator). ``` ## 为什么这些规则 - **每页唯一的标题和描述**是内容网站最具影响力的SEO规则。批量生成内容页面的代理通常重复使用相同的元数据模板，生成数十个技术上不同但对爬虫而言看似相同的页面——从而触发软去重惩罚。 - **每个页面都设置规范URL，而不仅仅是重复页面**经常被误解。阅读SEO文档的代理通常只在内容明显重复的地方添加规范URL（例如分页）。实际上，每个页面都应自引用其规范URL，以防止参数注入的爬取变体分散链接权益。 ## 适用场景 - 博客、文档网站、SEO内容中心和营销网站，这些网站以自然搜索为主要获取渠道。 ## 不适用 - 内部工具、仪表盘或应用程序，其中SEO无关紧要——规范URL/结构化数据要求会增加无益的开销。