` and `<meta name="description">` from every built page and count duplicates. Zero duplicate titles is the target. ## 6. Social cards - Open Graph `title`, `description`, `url`, `siteName`, `type` and an image of 1200×630 (`app/opengraph-image.tsx` with `next/og`), plus `twitter: { card: "summary_large_image" }`. - **Verify** the rendered `<meta property="og:image">` is an absolute production URL and that the image returns 200 with `content-type: image/png`. ## 7. Structured data - Add only what is true: `WebSite` and `Organization` on the home page, `BreadcrumbList` on nested pages, an item type that honestly fits (e.g. `SoftwareSourceCode`, `CreativeWork`, `Article`, `Product` only if it's for sale). - Never invent ratings, reviews, prices or FAQ markup for content that isn't a visible FAQ. - Render it as `<script type="application/ld+json">` with `<` escaped. **Verify** by parsing every block from the built HTML with `JSON.parse` and checking required fields; then spot-check in a rich-results validator. ## 8. Rendering and status codes - Important content is in the server-rendered HTML (`curl` the page and grep for the `h1` and body text). - Unknown slugs return a real 404 (`notFound()`), not a 200 with an empty state. - **Crawl** internal links from the home page and report every non-200 and every redirect chain. ## 9. URL structure - Lowercase, hyphenated, readable, stable: `/components/silk-field`, not `/c?id=42`. Changing a URL needs a 301 from the old one. ## Output A checklist with pass/fail and evidence (the command and what it printed) for each section, then the fixes, then the same checklist re-run.

--- name: technical-seo-audit description: Audit whether search engines can crawl, render, understand and correctly index a site - robots.txt, sitemap, canonical URLs, metadata, structured data, Open Graph and X cards, redirects, broken links, duplicate content, URL structure and programmatic pages - with Next.js App Router specifics. Use before launch, before a domain move, or when pages are missing from the index. --- # /technical-seo-audit Verify everything against the **production build** (`next build && next start`, or the deployed site), never the dev server. For page content and headings, run `/seo-auditor`. Don't promise rankings; make indexing correct. ## 1. One origin - Decide the production origin (e.g. `https://example.com`) and put it in one config value. - In Next.js set `metadataBase: new URL(origin)` in the root layout, so relative canonical and OG URLs resolve absolutely. - **Grep** the repo for `localhost`, old domains, placeholder domains and `http://`. None may reach production metadata, sitemaps or JSON-LD. - Redirect the other variants (http → https, www ↔ apex) with a single 301 hop. ## 2. robots.txt - `app/robots.ts` returning `{ rules, sitemap }`. Allow the site; disallow only what is genuinely not for search (iframe-only preview routes, internal tools, search-result URLs). - Never use robots.txt to remove a page from the index; that needs `noindex` on a crawlable page. - **Verify:** `curl $ORIGIN/robots.txt` shows the sitemap line with the absolute production URL. ## 3. Sitemap - `app/sitemap.ts` built from the same data as the routes (the registry, the CMS, the database), so it can't drift. - Include only canonical, indexable, 200-status URLs. No query-string variants, no noindexed pages, no redirects. - **Verify:** fetch every URL in the sitemap and assert 200 and that its canonical equals itself. ## 4. Canonicals and duplicates - Every indexable page sets `alternates: { canonical: path }`. - Filtered and sorted views (`?category=`, `?q=`, `?sort=`) canonicalise to the unfiltered page, or are noindexed. - Preview, print and embed routes: `robots: { index: false }`. - Trailing slashes and casing are consistent; duplicates 301 to one form. ## 5. Metadata on generated routes - Use `generateMetadata` on dynamic routes and build title, description and canonical from the item's own data. - **Check uniqueness:** extract `` and `<meta name="description">` from every built page and count duplicates. Zero duplicate titles is the target. ## 6. Social cards - Open Graph `title`, `description`, `url`, `siteName`, `type` and an image of 1200×630 (`app/opengraph-image.tsx` with `next/og`), plus `twitter: { card: "summary_large_image" }`. - **Verify** the rendered `<meta property="og:image">` is an absolute production URL and that the image returns 200 with `content-type: image/png`. ## 7. Structured data - Add only what is true: `WebSite` and `Organization` on the home page, `BreadcrumbList` on nested pages, an item type that honestly fits (e.g. `SoftwareSourceCode`, `CreativeWork`, `Article`, `Product` only if it's for sale). - Never invent ratings, reviews, prices or FAQ markup for content that isn't a visible FAQ. - Render it as `<script type="application/ld+json">` with `<` escaped. **Verify** by parsing every block from the built HTML with `JSON.parse` and checking required fields; then spot-check in a rich-results validator. ## 8. Rendering and status codes - Important content is in the server-rendered HTML (`curl` the page and grep for the `h1` and body text). - Unknown slugs return a real 404 (`notFound()`), not a 200 with an empty state. - **Crawl** internal links from the home page and report every non-200 and every redirect chain. ## 9. URL structure - Lowercase, hyphenated, readable, stable: `/components/silk-field`, not `/c?id=42`. Changing a URL needs a 301 from the old one. ## Output A checklist with pass/fail and evidence (the command and what it printed) for each section, then the fixes, then the same checklist re-run.