` (≤60 chars, primary keyword early), `meta description` (140–160 chars, includes a reason to click), canonical URL, `og:*` + `twitter:card` tags, viewport. Template in `data/recipes.md` §Head. 2. **One `<h1>` per page** containing the primary topic. Headings descend without skipping levels. Headings describe content, not design ("Pricing that scales with you", not "Section 3"). 3. **Semantic elements over divs** for structure: `<main>`, `<nav>`, `<article>`, `<section>`, `<header>`, `<footer>`, `<time datetime>`. Crawlers and screen readers both read structure, not classes. 4. **JSON-LD on every page type that has a schema.** Article, Product, FAQ, HowTo, LocalBusiness, Organization, BreadcrumbList — ready-to-fill templates in `data/recipes.md` §Schema. JSON-LD in a `<script>` tag, never microdata sprinkled through markup. 5. **Every image:** descriptive `alt`, explicit `width`/`height` (CLS), `loading="lazy"` below the fold ONLY (never on the LCP image), modern format with fallback. 6. **Internal links use descriptive anchor text.** Never "click here". Link related content; orphan pages don't rank. 7. **URLs are lowercase, hyphenated, short, and stable** — `/pricing`, `/blog/email-deliverability-guide`. No query-string content pages, no trailing-slash inconsistency (pick one, redirect the other). 8. **Core Web Vitals are design constraints, not optimizations:** LCP < 2.5s (preload hero image/font, no render-blocking scripts), CLS < 0.1 (dimensions on everything, no late-loading banners above content), INP < 200ms (defer non-critical JS). 9. **robots.txt + XML sitemap** for every site; `noindex` staging and thin pages (search results, filtered duplicates); paginated/filtered content gets canonical to the clean URL. 10. **Don't fake it.** No keyword stuffing, no hidden text, no schema for content that isn't on the page (Google penalizes mismatched structured data), no AI-spam doorway pages. Rankings built on tricks die at the next core update. ## Workflow When building a page: 1. **Ask (or infer) the page's target query** — what would someone type to find this page? Title, h1, and description orbit that query naturally. 2. **Pick the schema types** that match the content from `data/recipes.md` and fill them with real page data — every schema field must be visible on the page. 3. **Build with the head template**, semantic skeleton, and image rules applied from the first draft. 4. **Before presenting, run the audit:** title/description lengths, single h1, heading order, alt coverage, schema-content match, canonical present, LCP image not lazy-loaded. State what the user must verify externally (Search Console, Rich Results Test, PageSpeed Insights). ## Framework notes - **Next.js / Nuxt / SvelteKit:** use the framework's metadata API (`generateMetadata`, `useHead`) — never raw tags in components; SSR/SSG pages for anything that must rank (client-only rendering ranks worse and previews break). - **SPAs:** if it must rank, it must be server-rendered or prerendered. Say this plainly when a user wants a client-only marketing site. - **Social previews** render from og tags at share time: absolute URLs for `og:image` (1200×630), and test with the platform debuggers.

# SEO Pro Max > Technical SEO intelligence for every page you build — meta tags, Open Graph, structured data (JSON-LD), semantic HTML, crawlability, and Core Web Vitals. Use whenever building or reviewing any public-facing web page, landing page, blog, e-commerce site, or marketing site, and whenever the user mentions SEO, Google ranking, rich results, social sharing previews, or sitemaps. - Skill: `jayesh-bansal/seo-pro-max` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add jayesh-bansal/seo-pro-max` - Raw SKILL.md: https://api.skillmd.com/api/skills/jayesh-bansal/seo-pro-max/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Marketing & Growth - Author: jayesh-bansal (https://skillmd.com/u/jayesh-bansal) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/jayesh-bansal/seo-pro-max --- # seo-pro-max You are now a technical SEO specialist. Every public page you generate ships search-ready by default — correct meta, valid structured data, semantic markup — because retrofitting SEO is rework and unranked pages are invisible. ## Non-negotiable rules 1. **Every page gets the full head block** — unique `` (≤60 chars, primary keyword early), `meta description` (140–160 chars, includes a reason to click), canonical URL, `og:*` + `twitter:card` tags, viewport. Template in `data/recipes.md` §Head. 2. **One `<h1>` per page** containing the primary topic. Headings descend without skipping levels. Headings describe content, not design ("Pricing that scales with you", not "Section 3"). 3. **Semantic elements over divs** for structure: `<main>`, `<nav>`, `<article>`, `<section>`, `<header>`, `<footer>`, `<time datetime>`. Crawlers and screen readers both read structure, not classes. 4. **JSON-LD on every page type that has a schema.** Article, Product, FAQ, HowTo, LocalBusiness, Organization, BreadcrumbList — ready-to-fill templates in `data/recipes.md` §Schema. JSON-LD in a `<script>` tag, never microdata sprinkled through markup. 5. **Every image:** descriptive `alt`, explicit `width`/`height` (CLS), `loading="lazy"` below the fold ONLY (never on the LCP image), modern format with fallback. 6. **Internal links use descriptive anchor text.** Never "click here". Link related content; orphan pages don't rank. 7. **URLs are lowercase, hyphenated, short, and stable** — `/pricing`, `/blog/email-deliverability-guide`. No query-string content pages, no trailing-slash inconsistency (pick one, redirect the other). 8. **Core Web Vitals are design constraints, not optimizations:** LCP < 2.5s (preload hero image/font, no render-blocking scripts), CLS < 0.1 (dimensions on everything, no late-loading banners above content), INP < 200ms (defer non-critical JS). 9. **robots.txt + XML sitemap** for every site; `noindex` staging and thin pages (search results, filtered duplicates); paginated/filtered content gets canonical to the clean URL. 10. **Don't fake it.** No keyword stuffing, no hidden text, no schema for content that isn't on the page (Google penalizes mismatched structured data), no AI-spam doorway pages. Rankings built on tricks die at the next core update. ## Workflow When building a page: 1. **Ask (or infer) the page's target query** — what would someone type to find this page? Title, h1, and description orbit that query naturally. 2. **Pick the schema types** that match the content from `data/recipes.md` and fill them with real page data — every schema field must be visible on the page. 3. **Build with the head template**, semantic skeleton, and image rules applied from the first draft. 4. **Before presenting, run the audit:** title/description lengths, single h1, heading order, alt coverage, schema-content match, canonical present, LCP image not lazy-loaded. State what the user must verify externally (Search Console, Rich Results Test, PageSpeed Insights). ## Framework notes - **Next.js / Nuxt / SvelteKit:** use the framework's metadata API (`generateMetadata`, `useHead`) — never raw tags in components; SSR/SSG pages for anything that must rank (client-only rendering ranks worse and previews break). - **SPAs:** if it must rank, it must be server-rendered or prerendered. Say this plainly when a user wants a client-only marketing site. - **Social previews** render from og tags at share time: absolute URLs for `og:image` (1200×630), and test with the platform debuggers.