Schema

evals/evals.json

{
  "skill_name": "schema",
  "evals": [
    {
      "id": 1,
      "prompt": "Add schema markup to our SaaS product's homepage. We're a project management tool called TaskFlow. We need Organization schema and any other relevant types.",
      "expected_output": "Should check for product-marketing.md first. Should implement Organization schema in JSON-LD format with all required and recommended properties (name, url, logo, description, sameAs for social profiles). Should recommend additional schema types for a SaaS homepage: WebSite (with SearchAction if applicable), SoftwareApplication or Product. Should use @graph for multiple schema types on one page. Should provide the complete JSON-LD code ready to implement. Should recommend validation with Google's Rich Results Test and Schema.org validator.",
      "assertions": [
        "Checks for product-marketing.md",
        "Implements Organization schema in JSON-LD",
        "Includes required and recommended properties",
        "Recommends additional relevant schema types",
        "Uses @graph for multiple types",
        "Provides complete JSON-LD code",
        "Recommends validation tools"
      ],
      "files": []
    },
    {
      "id": 2,
      "prompt": "We have a FAQ page with 20 questions about our product. How do I add FAQ schema to get the rich results in Google?",
      "expected_output": "Should implement FAQPage schema in JSON-LD format. Should show the correct structure: FAQPage as mainEntity containing Question items, each with acceptedAnswer. Should provide a complete code example with 2-3 sample questions. Should explain that FAQ schema can enable rich results showing questions/answers directly in search. Should note Google's guidelines for FAQ schema (factual answers, not promotional). Should recommend validation approach.",
      "assertions": [
        "Implements FAQPage schema in JSON-LD",
        "Shows correct nested structure (FAQPage > Question > Answer)",
        "Provides complete code example",
        "Explains rich result benefits",
        "Notes Google's FAQ schema guidelines",
        "Recommends validation"
      ],
      "files": []
    },
    {
      "id": 3,
      "prompt": "add schema to our blog posts. we publish articles about marketing tips.",
      "expected_output": "Should trigger on casual phrasing. Should implement Article (or BlogPosting) schema in JSON-LD. Should include required properties: headline, author (as Person with name and url), datePublished, dateModified, image, publisher (as Organization). Should recommend BreadcrumbList schema alongside the article schema. Should provide template code that can be reused across blog posts. Should address how to populate dynamic fields (date, author, headline) from the CMS.",
      "assertions": [
        "Triggers on casual phrasing",
        "Implements Article or BlogPosting schema",
        "Includes author, datePublished, image, publisher",
        "Recommends BreadcrumbList alongside",
        "Provides reusable template code",
        "Addresses CMS integration for dynamic fields"
      ],
      "files": []
    },
    {
      "id": 4,
      "prompt": "We're an e-commerce site selling physical products. What schema markup do we need for our product pages?",
      "expected_output": "Should implement Product schema with full properties: name, description, image, brand, sku, offers (with price, priceCurrency, availability, url). Should recommend AggregateRating if they have reviews, and Review schema for individual reviews. Should include BreadcrumbList for navigation. Should address common e-commerce schema types: Product, Offer, AggregateRating, Review. Should provide complete JSON-LD code. Should note that Product schema can enable rich results (price, availability, ratings in search).",
      "assertions": [
        "Implements Product schema with full properties",
        "Includes Offer with price, availability",
        "Recommends AggregateRating and Review schema",
        "Includes BreadcrumbList",
        "Provides complete JSON-LD code",
        "Notes rich result benefits for products"
      ],
      "files": []
    },
    {
      "id": 5,
      "prompt": "We added schema markup to our site but it's not showing rich results in Google. Can you help debug?",
      "expected_output": "Should provide a systematic debugging approach: first validate with Google Rich Results Test and Schema.org validator (syntax errors), then check for common issues (incorrect nesting, missing required properties, JSON-LD placement errors). Should explain that valid schema doesn't guarantee rich results — Google chooses when to show them. Should recommend checking Search Console for structured data reports and errors. Should address common debugging scenarios: schema not detected, warnings vs errors, eligible vs displayed.",
      "assertions": [
        "Recommends validation tools for debugging",
        "Checks for common schema errors",
        "Explains valid schema doesn't guarantee rich results",
        "Recommends Search Console structured data reports",
        "Addresses warnings vs errors distinction",
        "Provides systematic debugging approach"
      ],
      "files": []
    },
    {
      "id": 6,
      "prompt": "Our organic search traffic dropped after a site redesign. Can you do a technical SEO audit?",
      "expected_output": "Should recognize this is a technical SEO audit request, not a schema markup task. Should defer to or cross-reference the seo-audit skill, which handles comprehensive technical SEO audits. Schema markup is one component of SEO but doesn't address the broader technical issues (redirects, crawlability, indexation) that likely caused the traffic drop.",
      "assertions": [
        "Recognizes this as a technical SEO audit request",
        "References or defers to seo-audit skill",
        "Does not attempt full SEO audit using schema markup patterns"
      ],
      "files": []
    },
    {
      "id": 7,
      "prompt": "Our Next.js product page embeds JSON-LD with dangerouslySetInnerHTML and JSON.stringify(schema). Product names come from a CMS and can contain HTML, quotes, ampersands and Unicode. How do we keep product text intact without allowing it to terminate the JSON-LD script element?",
      "expected_output": "Should use the schema implementation example and escape the serialized JSON for its HTML script context by replacing every less-than character with a literal JSON Unicode escape before assigning __html. Should explain that JSON.stringify alone does not protect the HTML parser boundary, preserve the original product text when JSON is parsed, and avoid deleting content, HTML-entity escaping JSON or directly interpolating raw names. Should refer to the current official Next.js guidance.",
      "assertions": [
        "Identifies the script element HTML parsing boundary",
        "Escapes every less-than character after JSON serialization",
        "Preserves original product text after JSON.parse",
        "Does not treat JSON.stringify alone as safe HTML embedding",
        "Uses the linked Next.js guidance"
      ],
      "files": []
    },
    {
      "id": 8,
      "prompt": "It's October 2026. Our product FAQ has 20 questions. Add FAQ schema so Google shows expandable questions in our search result, and tell us how to validate that feature.",
      "expected_output": "Should explain before offering implementation that Google stopped showing FAQ rich results in May 2026, citing Google's current documentation updates. The earlier government/health exception does not apply now. Should separate Schema.org vocabulary validity from Google feature eligibility. If semantic FAQPage markup is still wanted for another consumer, ask for the actual visible questions and answers and use mainEntity > Question > acceptedAnswer; don't invent product answers. Recommend Schema.org Validator for vocabulary. Missing FAQ detection or reports in Rich Results Test or Search Console is not an implementation defect. Do not promise expandable results or tell the user to delete unrelated supported markup.",
      "assertions": [
        "Explains Google FAQ rich results were retired before proposing code",
        "Cites current official Google feature guidance",
        "Does not present the old government/health exception as current eligibility",
        "Separates Schema.org validity from Google feature eligibility",
        "Offers semantic markup only with a separate purpose and actual visible content",
        "Does not treat missing retired-feature detection/reporting as a defect",
        "Does not promise the requested retired search appearance"
      ],
      "files": []
    },
    {
      "id": 9,
      "prompt": "Our tutorial uses valid HowTo JSON-LD but the Rich Results Test does not detect it. Should we fix the syntax until Google shows tutorial steps? Also, our homepage lost its sitelinks search box; should we delete all WebSite markup?",
      "expected_output": "Should check current feature support and explain Google's HowTo and sitelinks search box retirements with official references. Valid vocabulary cannot restore those appearances. Use Schema.org Validator if there is a semantic consumer; do not diagnose missing retired-feature detection as a syntax failure. Preserve useful WebSite site-name markup and unrelated supported features. Separate remaining syntax/content issues from feature availability.",
      "assertions": [
        "Recognizes retired HowTo and sitelinks search box features",
        "Separates syntax errors from unavailable appearances",
        "Keeps WebSite site identity and unrelated supported markup",
        "Uses official feature-support sources and appropriate validators"
      ],
      "files": []
    }
  ]
}