SEO Page Builder
Every page your agent ships comes out technically SEO-complete.
An agent skill encoding technical SEO into page creation: metadata and canonical rules, heading hierarchy, structured data selection, internal linking, and Core Web Vitals habits. Your agent ships pages that are SEO-complete by construction instead of retrofitted.
SKILL.md
--- name: seo-page description: Technical SEO checklist for web pages. Use when creating or significantly modifying public-facing pages. --- # SEO Page Builder Apply to every public page created or restructured. ## Metadata (every page) - `<title>`: under 60 chars, primary phrase near the front, unique across the site - Meta description: 140-160 chars, contains the phrase naturally, written as the search-result pitch - Canonical URL set; parameterized/filtered variants canonicalize to the base - OpenGraph + Twitter card tags (title, description, image) ## Structure - Exactly one `<h1>`, containing the primary phrase - Heading hierarchy never skips levels; H2s should answer the questions searchers ask - Semantic elements (`<main>`, `<article>`, `<nav>`) over div soup - Images: descriptive alt text (not keyword stuffing), width/height attributes set ## Structured data Choose by page type and validate the JSON-LD: - Article/guide → Article · FAQ section → FAQPage · Product/tool → Product or SoftwareApplication - List page → ItemList · How-to content → HowTo · Breadcrumbs → BreadcrumbList everywhere ## Linking - Every new page: linked FROM at least 2 existing relevant pages (orphan pages don't rank) - Descriptive anchor text — never "click here" - New page added to the sitemap (verify the generator picks it up) ## Performance habits - Images lazy-loaded below the fold, priority-loaded for LCP element - No layout shift: dimensions reserved for images/embeds/ads - Fonts: display=swap, preload the primary face ## Do not - Generate doorway pages (near-duplicate pages targeting keyword variants) - Hide text or stuff keywords — write for the reader, mark up for the crawler
Try it out
Open in Playground →Run this skill on a real model without leaving the page. Every run is saved to your history for this skill.
How to use
- 1Save as .claude/skills/seo-page/SKILL.md — it activates when the agent builds public pages.
- 2For programmatic/pSEO pages, the linking and canonical rules matter most; review those in the PR.
- 3Validate structured data with Google's Rich Results test after significant page-type changes.
Examples
Agent building a comparison page
Input
Asked to add a '/compare/x-vs-y' page to a Next.js site.
Output
Page shipped with: unique title (52 chars), meta description, canonical, Article + BreadcrumbList JSON-LD, one h1, FAQ section with FAQPage markup, links added from /blog/x-guide and /alternatives hub, sitemap entry verified in build output.
Pro tips
- The 'linked from 2 existing pages' rule is the most commonly skipped and the most costly to skip.
Frequently asked questions
Does structured data actually affect rankings?+
It affects how your results display (rich snippets, FAQ dropdowns), which affects click-through — and eligibility requires valid markup. It's the highest-leverage low-effort item on the list.
How does the doorway-page rule apply to programmatic SEO?+
pSEO is legitimate when each page has genuinely distinct, useful content for its query. The rule blocks near-duplicates with swapped keywords — the pattern that gets sites penalized.