Skip to content

Latest commit

 

History

History
137 lines (102 loc) · 4.61 KB

SKILL.md

File metadata and controls

137 lines (102 loc) · 4.61 KB
name description license metadata
schema-json-optimization
Validates existing JSON-LD structured data and suggests relevant schemas based on auto-detected page type. Does NOT enforce all schema types — only checks what is applicable to the current page context. Use when auditing or generating structured data for search engine rich results.
MIT

Schema JSON Optimization

Validate existing JSON-LD and suggest only relevant schemas based on the detected page type.

Important: Do NOT blindly check for all schema types. First detect what kind of page this is, then only validate/suggest schemas that are applicable.

Step 1: Find and validate existing JSON-LD

Look for all <script type="application/ld+json"> elements:

  • None found → ⚠️ WARN "No JSON-LD schema markup found"
  • Found → ✅ PASS "Found N JSON-LD block(s)" → validate each:

Validation rules

  • Invalid JSON → ❌ FAIL "Could not parse JSON-LD"
  • Missing @context → ❌ FAIL
  • Missing @type → ❌ FAIL
  • Valid → ✅ PASS and check type-specific required fields (see Step 3)

Step 2: Auto-detect page type

Scan the HTML for signals to determine what type of page this is. A page may match zero, one, or multiple types:

Page type Detection signals
Article / Blog <meta property="article:published_time">, <article> tag, blog-like URL patterns (/blog/, /post/, /news/)
LocalBusiness <address> tag, <a href="tel:...">, [itemprop="telephone"], [itemprop="address"], contact page patterns
Product <meta property="og:type" content="product">, price elements ([itemprop="price"]), add-to-cart buttons
FAQ Multiple <details>/<summary> or FAQ-patterned heading + paragraph pairs
Organization About page with company info, <meta property="og:type" content="website"> on homepage
BreadcrumbList <nav aria-label="breadcrumb">, .breadcrumb class, <ol> with breadcrumb-like links
WebSite Homepage with site-wide search form

If no type is detected → only validate existing schemas, do not suggest new ones.

Step 3: Validate and suggest ONLY for detected types

Only if Article/Blog detected:

Check existing Article schema for:

Field Missing =
headline ⚠️ WARN
datePublished ⚠️ WARN
author ⚠️ WARN
image ⚠️ WARN
description ⚠️ WARN

If no Article schema exists, suggest generating one:

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "[from og:title, <title>, or <h1>]",
  "datePublished": "[from article:published_time or <time datetime>]",
  "author": { "@type": "Person", "name": "[from meta author]" },
  "image": "[from og:image or first article img]",
  "description": "[from meta description]",
  "url": "[from canonical or og:url]"
}

Only if LocalBusiness detected:

Check for:

Field Missing =
name ❌ FAIL
address ⚠️ WARN
telephone ⚠️ WARN

If no LocalBusiness schema exists, suggest generating one:

{
  "@context": "https://schema.org",
  "@type": "LocalBusiness",
  "name": "[from og:site_name or <title>]",
  "address": "[from <address> or itemprop=address]",
  "telephone": "[from itemprop=telephone or tel: link]",
  "url": "[from canonical or og:url]",
  "image": "[from og:image]"
}

Only if Product detected:

Check for:

Field Missing =
name ❌ FAIL
offers (with price and priceCurrency) ⚠️ WARN
image ⚠️ WARN
description ⚠️ WARN

Only if FAQ detected:

Suggest FAQPage schema with Question/Answer pairs extracted from the page.

Only if BreadcrumbList detected:

Suggest BreadcrumbList schema from the breadcrumb navigation structure.

Key principle

DO: "This page looks like a blog post → checking for Article schema"
DON'T: "Missing LocalBusiness schema" on a blog post page

The goal is relevance, not completeness. Only flag missing schemas that would actually benefit the detected page type.

Edge cases

  • A page may have multiple valid types (e.g. a product page with FAQ section → suggest both Product and FAQPage)
  • A page may have existing schemas that are valid but don't match the detected type — that's fine, still validate them
  • If detection is ambiguous, report what was detected and let the user decide
  • Schema @type can be an array (e.g. ["Article", "NewsArticle"])