<tasklist_context> Use TaskList tool to check for existing tasks related to this work.

If a related task exists, note its ID and mark it in_progress with TaskUpdate when starting. </tasklist_context>

<rules_context> Read SEO rules NOW:

Use Read tool: ${CLAUDE_PLUGIN_ROOT}/rules/seo.md

This defines page classification (marketing vs app), all required vitals, and configuration requirements. The skill goes deeper than these rules, but they are the shared baseline. </rules_context>

<progress_context> Use Read tool: docs/progress.md (first 50 lines)

Check for recent changes that might affect SEO (new pages, redesigns, migrations). </progress_context>

SEO Audit Workflow

Deep SEO audit for web projects. Analyzes codebase for technical SEO compliance, content optimization, and social sharing readiness. Optionally validates against a live site.

Process

Phase 1: Detect & Classify

Step 1: Detect Framework

Use Glob + Grep to detect project type:

Check	Tool	Pattern
Next.js	Grep	"next" in package.json
Remix	Grep	"@remix-run" in package.json
Astro	Grep	"astro" in package.json
SvelteKit	Grep	"@sveltejs/kit" in package.json
Nuxt	Grep	"nuxt" in package.json
Static HTML	Glob	*.html in root or public/

Record framework — this determines where to look for routes, meta tags, and config.

Step 2: Find All Routes/Pages

Framework-specific route discovery:

Framework	Glob Pattern	Route Pattern
Next.js (App Router)	app/**/page.{tsx,jsx,ts,js}	Directory = route
Next.js (Pages Router)	pages/**/*.{tsx,jsx,ts,js}	File = route
Remix	app/routes/**/*.{tsx,jsx,ts,js}	File = route
Astro	src/pages/**/*.{astro,md,mdx}	File = route
SvelteKit	src/routes/**/+page.svelte	Directory = route
Static HTML	**/*.html	File = route

Present discovered routes to user.

Step 3: Classify Pages

Ask the user to classify pages:

Present the route list and ask which are app-only (authenticated/gated). Use AskUserQuestion with multiSelect.

Default: treat all as marketing unless user marks as app-only.

Example:

"I found these routes. Which are app pages (authenticated/gated)?
Marketing pages get full SEO audit. App pages get basics only (title, noindex)."

Routes:
- / (homepage)
- /about
- /pricing
- /blog
- /blog/[slug]
- /dashboard
- /settings
- /api/*

API routes are automatically excluded from SEO checks.

Step 4: Check Existing SEO Config

Scan for what's already in place:

Check	Where to Look
robots.txt	public/robots.txt, app/robots.ts (Next.js)
sitemap	public/sitemap.xml, app/sitemap.ts (Next.js)
Meta setup	Root layout, page-level metadata exports
Structured data	JSON-LD in layouts or pages
OG images	app/opengraph-image.*, static OG images in public/
Favicons	app/icon.*, app/favicon.ico, public/favicon.ico
Google Search Console	<meta name="google-site-verification"> in root layout, public/google*.html verification file

Report what exists:

## Existing SEO Config
- ✓ robots.txt present
- ✓ Meta titles in root layout
- ✗ sitemap.xml missing
- ✗ No structured data found
- ✗ OG images not configured
- ✗ Google Search Console not verified

Step 5: Ask for Live URL

"Do you have a live URL (dev or production) for this site? If so, I can run Lighthouse and PageSpeed for additional analysis. This is optional."

If provided, store for Phase 2.

Phase 2: Audit

Run checks against the codebase, organized by category. Marketing pages get full treatment. App pages get basics only.

Category 1: Crawlability

• robots.txt — Present? Any marketing paths blocked? Sitemap referenced?
• Meta robots — Any marketing pages with noindex? Common pitfall: Vercel preview noindex leaking to production.
• Sitemap — Present? Lists all marketing pages? Doesn't list app pages? Dynamically generated or static?
• Redirect chains — Any 301 → 301 → page chains? (Check middleware, vercel.json, next.config redirects)
• Google Search Console — Site verification detected? (meta tag google-site-verification or public/google*.html file). If not verified, flag as High — without Search Console, Google won't notify you of indexing issues, manual actions, or crawl errors.

Category 2: Indexability

• Canonical tags — Present on every marketing page? Absolute URLs? Self-referencing where appropriate?
• Duplicate content — Same title or description on multiple pages? Multiple URLs serving identical content?
• Hreflang — If i18n detected (next-intl, i18next, locale folders): hreflang tags present? Return links correct?

Category 3: On-Page

• Titles — Present? Unique per page? Under 60 chars? Not generic ("Home", "Page", "Untitled")? Descriptive of page content?
• Meta descriptions — Present? Unique per page? Under 160 chars? Not boilerplate? Includes call-to-action where appropriate?
• Heading hierarchy — Single h1 per page? Logical nesting (h1 → h2 → h3)? No skipped levels? h1 content meaningful?
• Image alt text — All meaningful images have descriptive alt? Decorative images use alt=""? No generic alt ("image", "photo")?
• URL structure — Clean, readable URLs? No UUIDs? No excessive nesting? No query params for content pages?

Category 4: Structured Data

• Presence — JSON-LD blocks present on key page types?
• Schema types — Appropriate for content?
  • Homepage: Organization or WebSite
  • Blog posts: Article or BlogPosting
  • Product pages: Product
  • FAQ pages: FAQPage
  • About: Organization or Person
• Coverage gaps — Some page types have structured data but others don't?
• Validity — Required properties present for each schema type?

Category 5: Social & Meta

• Open Graph — og:title, og:description, og:image set on all marketing pages?
• OG image — 1200x630 dimensions? Exists and loads?
• Twitter Cards — twitter:card type set (summary or summary_large_image)? twitter:title, twitter:description, twitter:image present?
• Consistency — OG title/description match or complement the page title/description?

Category 6: Technical Foundations

• Language — <html lang="..."> set to correct language code?
• Viewport — <meta name="viewport" content="width=device-width, initial-scale=1"> present?
• Charset — <meta charset="utf-8"> or equivalent declared?
• HTTPS — Site enforces HTTPS? No mixed content?

Live Site Checks (Optional)

If user provided a live URL and wants live checks:

"Running Lighthouse and PageSpeed on [URL]. This may take a moment."

Run programmatic tools (blocking — results go into the report):

# Lighthouse SEO audit
npx lighthouse [URL] --output=json --only-categories=seo --chrome-flags="--headless=new" 2>/dev/null

# Lighthouse Performance (Core Web Vitals)
npx lighthouse [URL] --output=json --only-categories=performance --chrome-flags="--headless=new" 2>/dev/null

PageSpeed Insights API (no key needed for light usage):

GET https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url=[URL]&category=seo&strategy=mobile
GET https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url=[URL]&category=seo&strategy=desktop

If user wants full-site crawl:

npx unlighthouse --site [URL] --reporter jsonExpanded

Failure handling: If any check fails (timeout, auth wall, network error), note in the report: "Could not access [URL] — [reason]. Skipping live check." Continue with codebase-only findings.

Extract from Lighthouse JSON:

• SEO score (0-100)
• Specific audit failures (missing meta descriptions, missing alt text, etc.)
• Core Web Vitals (LCP, CLS, INP)

Extract from PageSpeed API:

• Mobile and desktop SEO scores
• Opportunities for improvement

Phase 3: Report & Act

Step 1: Generate Report

Create: docs/audits/YYYY-MM-DD-seo-audit.md

# SEO Audit Report

**Date:** YYYY-MM-DD
**Framework:** [detected framework]
**Marketing pages:** [count]
**App pages:** [count] (basics only)
**Live URL:** [URL or "not provided"]
**Live checks:** [run / not run]

## Summary

[1-2 paragraph overview of findings]

- **Critical:** X issues
- **High:** X issues
- **Medium:** X issues

## Critical Issues

> Indexing is broken or severely impaired

### [Issue Title]
**File:** `path/to/file.ts:123`
**Category:** [Crawlability / Indexability / On-page / etc.]
**Issue:** [What's wrong]
**Impact:** [How this affects SEO]
**Fix:** [Specific change needed]

## High Priority

> Core SEO elements missing

[Same format per finding]

## Medium Priority

> Suboptimal but not broken

[Same format per finding]

## Live Site Results

> From Lighthouse and PageSpeed (if run)

**Lighthouse SEO Score:** [X/100]
**Core Web Vitals:**
- LCP: [value] — [good/needs improvement/poor]
- CLS: [value] — [good/needs improvement/poor]
- INP: [value] — [good/needs improvement/poor]

**PageSpeed:**
- Mobile SEO: [X/100]
- Desktop SEO: [X/100]

[Specific audit failures from Lighthouse]

## Google Search Console

**Status:** [Verified / Not verified]

If not verified, submit your site now:
→ **Add your site:** [https://search.google.com/search-console/welcome](https://search.google.com/search-console/welcome)

After verification, submit your sitemap:
→ **Submit sitemap:** `https://search.google.com/search-console/sitemaps?resource_id=https://[DOMAIN]/`

Without Search Console, Google won't alert you to indexing problems, crawl errors, or manual actions. This is not optional for any site that needs organic traffic.

## Manual Validation Tools

Check these tools for additional validation:
- [Google Rich Results Test](https://search.google.com/test/rich-results?url=[URL])
- [Facebook Sharing Debugger](https://developers.facebook.com/tools/debug/?q=[URL])
- [Twitter Card Validator](https://cards-dev.twitter.com/validator)
- [LinkedIn Post Inspector](https://www.linkedin.com/post-inspector/)

## Framework-Specific Recommendations

[Tailored to detected framework — Next.js Metadata API, Astro SEO patterns, etc.]

Step 2: Determine Fix Approach

Count the number of files affected by findings.

Fix heuristic:

• 1-3 files affected: Offer to fix directly. "I found [N] issues across [N] files. I can fix these now. Should I?"
• 4-10 files affected: Give the user a choice. "There are [N] issues across [N] files. Want me to fix them now, or create an implementation plan with /arc:detail?"
• 10+ files affected: Recommend a plan. "There are [N] issues across [N] files. This needs a structured approach. Want me to create an implementation plan with /arc:detail?"

Use AskUserQuestion to present the appropriate options based on file count.

Step 3: Framework-Specific Advice

Tailor recommendations to the detected framework:

Next.js (App Router):

• Use Metadata API (export const metadata or generateMetadata)
• Use opengraph-image.tsx for dynamic OG images
• Use robots.ts for programmatic robots.txt
• Use sitemap.ts for dynamic sitemap generation
• Use icon.tsx for dynamic favicons

Next.js (Pages Router):

• Use next/head for meta tags
• Use next-seo package or custom SEO component
• Static sitemap generation with next-sitemap

Remix:

• Use meta function exports per route
• Handle OG images in resource routes

Astro:

• Use <head> in layout components
• SEO component pattern for reusable meta
• Built-in sitemap integration (@astrojs/sitemap)

Other:

• Standard HTML meta tag patterns
• Suggest popular SEO packages if applicable

Step 4: Commit Report

mkdir -p docs/audits
git add docs/audits/YYYY-MM-DD-seo-audit.md
git commit -m "docs: add SEO audit report"

Step 5: Present Summary & Next Steps

## SEO Audit Complete

**Scope:** [N] marketing pages, [N] app pages
**Live checks:** [Yes — score X/100 / No]
**Report:** docs/audits/YYYY-MM-DD-seo-audit.md

### Findings
- Critical: X | High: X | Medium: X
- Files affected: X

### [Fix option based on heuristic]

<success_criteria> SEO audit is complete when:

• Framework detected
• Routes discovered and classified (marketing vs app)
• Existing SEO config checked
• All 6 categories audited against marketing pages
• App pages checked for basics (title, noindex)
• Live site checks run (if opted in) and results in report
• Report generated in docs/audits/
• Report committed
• Fix approach offered (direct fix, plan, or done)
• Framework-specific recommendations included
• Manual validation tool links included (with pre-filled URLs if live URL provided)
• Google Search Console verification checked; if missing, user given direct link to submit
• Progress journal updated </success_criteria>

Interop

• Reads rules/seo.md for baseline vitals
• References /arc:detail for creating implementation plans (10+ file fixes)
• Can be invoked after /arc:letsgo for deeper analysis
• SEO agent (seo-engineer) handles the lighter audit in /arc:audit

<arc_log> After completing this skill, append to the activity log. See: ${CLAUDE_PLUGIN_ROOT}/references/arc-log.md

Entry: /arc:seo — [scope] audit ([N] issues) </arc_log>