Schema Detector

A Claude Skill that audits any URL for structured data — extracts every JSON-LD, Microdata, and RDFa block on the page, runs 12 schema-type checks, and returns a deterministic JSON report with severity-ranked findings (P0–P3) and 7-Lever framework attribution.

---

The name

Schema is what tells robots what your page is about. Most pages get it wrong, get it broken, or skip it entirely. This skill is the detector — point it at a URL, find out what's missing, get the fix.

Built by Ivan Nonveiller — enterprise SEO/GEO strategist.

---

What it does

You give it a URL. It returns a structured audit of every schema block on the page, plus a list of what's missing or broken.

• Fetches the target page (with web_search fallback for bot-blocked sites)
• Extracts JSON-LD, Microdata, and RDFa structured data
• Infers the page type (homepage, article, product, local-business, event, video, person)
• Runs 12 schema-type checks plus a meta-check for malformed JSON-LD
• Sorts findings by severity (P0 critical → P3 low)
• Maps every finding to one of the 7 Levers from the Ygramul methodology
• Returns: machine-readable JSON, chainable into other skills and audit workflows

---

Why this instead of Google's Rich Results Test

The Rich Results Test answers "is this valid for rich results today?" — useful, but limited. It tells you nothing about AI search citation, entity anchoring, or whether the schema you have matches the page type Google thinks you are.

This skill audits for what AI search engines and crawlers actually use, not just what currently triggers a SERP rich result. Findings include:

• Author as a string vs. linked Person entity (E-E-A-T signal that survived the 2023 deprecations)
• Missing sameAs arrays on Organization and Person schema (the disambiguation lifeline LLMs cross-reference)
• WebSite schema without SearchAction (still expected; sitelinks search box still matters)
• Page-type mismatches (Article schema on a product page is its own kind of broken)

It also flags FAQPage and HowTo schema as P3 with deprecation context — present in Rich Results Test as "valid" but no longer driving rich results for most sites since 2023.

---

The 12 checks

Lever 6 — Authority & GEO Visibility (entity anchors)

• Organization
• WebSite
• Person
• LocalBusiness

Lever 4 — Content Production & Optimization (page schemas)

• Article (incl. NewsArticle, BlogPosting)
• Product
• Review / AggregateRating
• Event
• VideoObject
• FAQPage (flagged P3 — deprecated for most sites)
• HowTo (flagged P3 — deprecated)

Lever 2 — Canonical Architecture

• BreadcrumbList

Lever 1 — Crawl & Indexation (meta-check)

• Invalid JSON-LD detection — surfaces malformed blocks as P0 before any other check runs

Severity is page-type-aware. Missing Product schema is P0 on a product page but P2 on a homepage. The full rule set lives in references/schema-checks.md.

---

Install

1. Download the contents of this repo (or clone it)
2. Open claude.ai → click your profile → Settings → Capabilities → Skills
3. Click Upload skill and select the SKILL.md file (the references load automatically)
4. Done — the skill is now available in your account

Requires Claude Pro, Team, or Enterprise. Free tier doesn't support custom skills.

---

Use

Start a new chat and type:

Audit the schema on https://yourwebsite.com/your-page

Or any natural variant — "what schema is on this page", "run schema detector on...", "check structured data for...".

Claude will:

1. Fetch the page
2. Extract all structured data
3. Run the 12 checks
4. Return JSON with detected schemas, findings sorted P0→P3, and a summary block

If you'd rather have a Markdown report than JSON, ask for it explicitly. The default is JSON because that's what chains into other tools.

---

Output shape

{
  "url": "https://example.com/blog/post",
  "fetchedAt": "2026-05-02T14:30:00.000Z",
  "pageTitle": "Why Sleep Matters",
  "pageType": "article",
  "fetchMethod": "web_fetch",
  "detectedSchemas": [
    { "type": "Article", "format": "json-ld", "rawTypeValue": "BlogPosting" },
    { "type": "Organization", "format": "json-ld", "rawTypeValue": "Organization" }
  ],
  "detectedTypeKeys": ["Organization", "Article"],
  "findings": [
    {
      "schemaType": "Article",
      "severity": "P0",
      "title": "Article missing required: datePublished, author",
      "detail": "Google requires headline, datePublished, and author for Article schema to be eligible for Top Stories and standard rich results.",
      "recommendation": "Add datePublished as ISO 8601, author as a Person object with @id and url.",
      "lever": "Lever 4 — Content Production & Optimization"
    }
  ],
  "summary": {
    "totalDetectedTypes": 2,
    "totalChecked": 12,
    "counts": { "P0": 1, "P1": 0, "P2": 2, "P3": 0 }
  }
}

---

Expected results

• Per-audit cost: ~$0.05–0.20 in Claude usage (covered by your Pro/Team/Enterprise plan)
• Audit time: 15–45 seconds depending on page size
• Output is deterministic — same URL produces the same finding shape every run, suitable for diffing audits over time

For schemas you fix:

• 1–2 weeks after shipping: Google re-crawls and re-validates
• 2–6 weeks: rich results begin appearing for newly-eligible schema (assuming Google approves)
• AI search citation effects (Perplexity, ChatGPT, AI Overviews) compound over months as entity signals strengthen

---

When not to use

• Sitewide audits — this skill is single-URL by design. For full-site work, use a dedicated technical SEO audit tool or a sitemap-aware audit skill.
• JS-rendered SPAs that inject schema client-side — web_fetch can't execute JavaScript. The skill will undercount and tag the audit with fetchMethod: "web_search_fallback" so you know.
• Pages still in draft — it needs the published version to read.
• Schema implementation — this skill audits and recommends. It doesn't write the JSON-LD for you. Pair with a separate content brief or implementation skill.

---

Companion skills

This is the focused single-page check. Two adjacent skills cover different scopes:

• ygramul-aio-cascade — Lever 4 AIO opener block generation. Run this after fixing schema gaps to capture AI Overview citations.
• tech-seo-audit — full-site technical audit covering all 7 Levers. Use when Schema Detector findings suggest deeper problems (3+ P0s, sitewide gaps).

---

Contributing

Found a check rule that needs tightening? Severity assignment that doesn't match real-world impact? PRs welcome. Open an issue first for anything bigger than a typo fix.

The check rules live in references/schema-checks.md — edit there. The skill is designed to be tuned without touching SKILL.md itself.

---

License

MIT. Use it, modify it, ship it in your own tools.

---

About the author

Ivan Nonveiller — enterprise technical SEO and GEO strategist. 10+ years optimizing search for brands including Walmart, AutoTRADER, Ferrari, Maserati, Ritz-Carlton, Sotheby's, and BetterSleep. Based in Montréal.

• LinkedIn: ivannonveiller
• Ygramul methodology: rushmediaagency.com

This skill is the open-source version of Ygramul's Lever 6 (Authority) and Lever 4 (Content) entity-schema check system.