Noir SEO Check

noir-seo-check — SEO + structured data for NOIR public pages

NOIR has significant public-facing surface area: blog posts, product detail pages, product category pages, public tenant landing. Each needs proper metadata to rank and share well. This skill walks a full audit + remediation for a given page type.

When to invoke

User says anything like:

• "audit SEO for the blog"
• "the product pages don't have Open Graph tags"
• "Google doesn't show rich snippets for our reviews"
• "generate structured data for X"
• "check meta descriptions"
• "fix sitemap"

Prerequisites

• The marketing-skills plugin (from coreyhaines31/marketingskills) is installed — it provides generic seo-audit + schema-markup workflows. This skill layers NOIR-specific knowledge on top.
• Reference doc: docs/backend/research/seo-meta-and-hint-text-best-practices.md — the team's SEO principles. Read this before audit.

Page types + what to check

Workflow

Phase 1 — Audit (read-only)

1. Run the generic audit first. If the seo-audit skill from marketing-skills is available, invoke it to get the baseline (broken links, missing meta, duplicate titles).
2. Check the NOIR-specific dimensions:
   • Every public page component has a <Helmet> or react-helmet-async block OR equivalent (meta tags rendered server-side for bot visibility — if client-rendered only, Google + social crawlers often can't read them)
   • Check SSR strategy: grep for react-helmet, Head, or <title> dynamic binding. If the frontend is pure CSR and there's no SSR/prerender layer, SEO is fundamentally broken — flag this as HIGH severity
   • Meta description in BlogPost / Product entity has been populated (seeder, admin editor, or auto-generated)
   • featuredImageUrl / thumbnailUrl is absolute (not relative) for og:image
   • slug column has a unique-per-tenant index (canonical URL correctness)
3. Check structured data:
   • Are JSON-LD <script type="application/ld+json"> blocks rendered?
   • Do they validate against schema.org? Test with the Rich Results Test tool
   • For products with reviews: is aggregateRating aggregated from ReviewDto?
4. Check sitemap:
   • GET /sitemap.xml returns valid XML (if NOIR has one)
   • Includes all public blog posts + products + categories
   • <lastmod> reflects entity ModifiedAt
   • Robots.txt points at the sitemap
5. Check Core Web Vitals impact:
   • Image dimensions set (prevents CLS)
   • Fonts preloaded
   • Critical CSS inlined (unless Vite's default + Tailwind coverage is sufficient)

Phase 2 — Report

Output a findings table:

Phase 3 — Remediation (with user consent)

For each HIGH/MEDIUM issue:

1. Identify the affected files (grep for the page component)
2. Show the diff before applying
3. Update BOTH the page component AND any relevant DTO/entity (e.g. add MetaTitle, MetaDescription, OgImageUrl columns to BlogPost via migration if missing)
4. If backfilling DB data: write a one-time seeder or Hangfire job, don't update rows manually
5. Add/update i18n keys if the meta is multi-locale (EN + VI)

Phase 4 — Verify

• View-source the page — all tags present
• Run the Rich Results Test on a representative URL
• Re-run audit-website / seo-audit — deltas should show only fixes
• Update docs/backend/research/seo-meta-and-hint-text-best-practices.md if a new pattern emerged

Anti-patterns this skill prevents

• Client-rendered meta tags — React setting <title> only on mount; Googlebot sees the pre-render HTML which has the default title. If not SSR'd or prerendered, you must either migrate the public routes to SSR, add a prerender service (Prerender.io, Rendertron), or generate static meta at build time.
• Relative og:image URLs — Facebook/Twitter/LinkedIn unfurl uses the origin from their crawler, not the page. Always absolute.
• Duplicate canonical URLs — tenant-slug pages that share a slug across tenants (e.g. acme/products/widget and globex/products/widget) need tenant-scoped canonicals. Rule 18 (tenant-scoped unique indexes) supports this at the DB layer; the canonical URL must reflect the tenant subdomain or path.
• Missing aggregateRating — you have ReviewDto data; aggregate it into Product structured data, don't leave it as separate Review blocks
• Over-stuffed keywords meta — Google ignores <meta name="keywords"> entirely; don't bother
• Boilerplate meta descriptions — "Welcome to our site" pulls zero click-through. Generate per-entity or require editors to write one
• Empty ModifiedAt — sitemap <lastmod> defaults to CreatedAt, signaling stale content. Use ModifiedAt from audit columns (Rule 11 + audit columns standard)

Cross-references

• marketing-skills plugin → seo-audit, schema-markup, site-architecture, programmatic-seo, analytics-tracking
• accessibility-compliance plugin → alt-text and ARIA issues overlap with SEO (image alt, heading hierarchy)
• ui-audit project skill → catches accessibility issues that also hurt SEO
• docs/backend/research/seo-meta-and-hint-text-best-practices.md — team's SEO principles (living doc)

Source: NOIR-Solution/NOIR — distributed by TomeVault.