--- context: fork name: schema-generator description: "Generates JSON-LD structured data, llms.txt, and robots.txt snippets for AEO/GEO (answer engine and generative engine optimization). Converts aeo-geo-2026 reference recommendations into executable templates. Use when user asks about schema markup, JSON-LD templates, structured data, AEO 실행, GEO 구현, llms.txt, citation capsule, FAQPage, HowTo, OpinionArticle, or 구조화 데이터." platforms: [claude-code, gemini-cli, codex-cli, cursor] level: 3 triggers: - "schema" - "JSON-LD" - "structured data" - "AEO 실행" - "GEO 구현" - "llms.txt" - "citation capsule" - "FAQPage" - "HowTo schema" - "OpinionArticle" agents: - "content-marketer" - "seo-specialist" tokens: "~5K" category: "marketing" --- # Schema Generator ## When This Skill Applies - Writing JSON-LD blocks for articles, FAQ pages, interviews, reviews, and tutorials - Shipping `llms.txt` so AI answer engines get a curated site map - Updating `robots.txt` to allow or deny specific AI crawler user agents - Wrapping a load-bearing passage as a citation capsule for answer-engine lift - Picking the right schema type when content does not fit the default `Article` - Validating structured data before publish using public validator URLs This skill is the executable counterpart to `skills/copywriting/references/aeo-geo-2026.md`. The reference explains *why*; this skill hands back templates, snippets, and a decision table. ## Placeholder Convention All templates in this skill and in `references/json-ld-schemas.md` use a single placeholder format: **``** — an `UPPER_SNAKE_CASE` token wrapped in angle brackets. | Rule | Example | |---|---| | Wrap every placeholder in `<` and `>` | `` | | Use `UPPER_SNAKE_CASE` for the inner name | `` | | Include unit or format hints in the name when helpful | ``, `` | | Strip every `` token before deploy | Find-and-replace pass with editor regex `<[A-Z0-9_]+>` | Editors can sweep for unresolved placeholders with a single regex: `<[A-Z0-9_]+>`. Any hit in a shipped JSON-LD file is a bug. ## Core Guidance ### 1. Schema Type Decision Table Pick one primary type per page. Add `BreadcrumbList` on every page regardless. | Content type | Primary schema | Secondary / nested | Reason | |---|---|---|---| | Blog post, essay | `Article` | `BreadcrumbList`, `Person` (author) | Default; Google News indexing, FAQ extraction possible | | Case study | `Article` | `Review` (for quantitative outcomes) | Structures outcome metrics so answer engines can lift them | | Opinion column | `OpinionArticle` (Article subtype) | `Person` | Marks subjective stance; answer engines treat it differently from reporting | | Interview feature | `Article` | `InterviewObject`, `Person` (interviewee, interviewer) | Carries both-party metadata | | Thought leadership | `Article` | `Person` (with `knowsAbout`, `alumniOf`) | Supplies E-E-A-T author authority signals | | FAQ section or page | `FAQPage` | — | AI engines pull Q&A directly; mandatory for FAQ-shaped content | | Tutorial, step-by-step | `HowTo` | `HowToStep`, `HowToTool` | Structures steps, tools, timings for answer boxes | | Product page | `Product` | `Offer`, `AggregateRating` | Out of scope for editorial skill; use e-commerce stack | | Homepage, brand hub | `Organization` | `WebSite`, `SearchAction` | Publisher identity + sitelinks search box | **Anti-pattern**: do not stack two primary types on the same page (e.g., `Article` + `HowTo` at top level). Nest one inside the other or split the page. ### 2. Required vs. Recommended Fields (quick reference) | Schema | Required | Strongly recommended | Optional but useful | |---|---|---|---| | `Article` | `headline`, `author`, `datePublished` | `image`, `dateModified`, `publisher`, `mainEntityOfPage` | `articleSection`, `keywords`, `wordCount` | | `OpinionArticle` | same as `Article` | `author.jobTitle`, `about` | `editor`, `backstory` | | `FAQPage` | `mainEntity` (array of `Question`) | each `Question.acceptedAnswer.text` | `about` topic linkage | | `HowTo` | `name`, `step` | `totalTime`, `estimatedCost`, `tool`, `supply` | `yield`, `prepTime`, `performTime` | | `Review` | `itemReviewed`, `reviewRating`, `author` | `reviewBody`, `datePublished` | `publisher` | | `Person` | `name` | `url`, `jobTitle`, `worksFor`, `sameAs` | `knowsAbout`, `alumniOf` | | `Organization` | `name`, `url` | `logo`, `sameAs` | `contactPoint`, `foundingDate` | | `BreadcrumbList` | `itemListElement` (ordered) | `position`, `name`, `item` | — | Full templates with all fields live in `references/json-ld-schemas.md`. ### 3. Minimal `Article` JSON-LD Template Placeholders follow the `` convention defined above — `UPPER_SNAKE_CASE` inside angle brackets, find-and-replace before deploy. ```json { "@context": "https://schema.org", "@type": "Article", "headline": "", "description": "", "image": ["", "", ""], "datePublished": "", "dateModified": "", "author": { "@type": "Person", "name": "", "url": "" }, "publisher": { "@type": "Organization", "name": "", "logo": { "@type": "ImageObject", "url": "" } }, "mainEntityOfPage": { "@type": "WebPage", "@id": "" } } ``` ### 4. `FAQPage` Template One script tag, one `FAQPage`, N `Question`s. Questions must match visible page text exactly — mismatches trigger manual action penalties. ```json { "@context": "https://schema.org", "@type": "FAQPage", "mainEntity": [ { "@type": "Question", "name": "", "acceptedAnswer": { "@type": "Answer", "text": "" } }, { "@type": "Question", "name": "", "acceptedAnswer": { "@type": "Answer", "text": "" } } ] } ``` ### 5. `HowTo` Template (for tutorials) ```json { "@context": "https://schema.org", "@type": "HowTo", "name": "", "description": "", "totalTime": "", "tool": [ { "@type": "HowToTool", "name": "" } ], "step": [ { "@type": "HowToStep", "position": 1, "name": "", "text": "", "image": "" }, { "@type": "HowToStep", "position": 2, "name": "", "text": "" } ] } ``` ### 6. `OpinionArticle` Template (for columns) `OpinionArticle` inherits from `Article`. The declared type signals to answer engines that the content is subjective, which changes how it is cited. ```json { "@context": "https://schema.org", "@type": "OpinionArticle", "headline": "", "author": { "@type": "Person", "name": "", "jobTitle": "", "worksFor": { "@type": "Organization", "name": "" } }, "datePublished": "", "dateModified": "", "about": "", "mainEntityOfPage": { "@type": "WebPage", "@id": "" } } ``` ### 7. Interview Feature (`Article` + `InterviewObject`) ```json { "@context": "https://schema.org", "@type": "Article", "headline": "", "datePublished": "", "author": [ { "@type": "Person", "name": "" } ], "mainEntity": { "@type": "InterviewObject", "interviewer": { "@type": "Person", "name": "" }, "interviewee": { "@type": "Person", "name": "", "jobTitle": "", "worksFor": { "@type": "Organization", "name": "" } } } } ``` > As of 2026-04, `InterviewObject` is an emerging schema.org type with limited AI engine support. Verify current adoption before production use. See json-ld-schemas.md:446 for extended note. ### 8. `BreadcrumbList` Template (pair with every primary schema) ```json { "@context": "https://schema.org", "@type": "BreadcrumbList", "itemListElement": [ { "@type": "ListItem", "position": 1, "name": "", "item": "" }, { "@type": "ListItem", "position": 2, "name": "", "item": "" }, { "@type": "ListItem", "position": 3, "name": "" } ] } ``` The final item intentionally omits `item` — it is the current page. ### 9. Citation Capsule Markup A citation capsule is the HTML wrapper around a citable passage (defined in `aeo-geo-2026.md`). Purpose: make retrieval systems recognize the passage as a self-contained quotable block. **HTML pattern**: ```html

Source: ,
``` **Markdown pattern** (for platforms without raw HTML): ```markdown > <120-180 word passage with proper noun, number, and source anchor> > > — , , . Source: ``` Place 2-4 capsules per 1500-word article, first one inside the first 500 words. ### 10. llms.txt and robots.txt Policy See `references/llms-txt-template.md` for the full `llms.txt` template and the robots.txt AI-crawler policy block. Quick summary: | File | Purpose | Status (as of 2026, verify current) | |---|---|---| | `robots.txt` | Allow/deny specific user agents | Long-standing standard; AI crawler UAs change — re-check vendor docs quarterly | | `llms.txt` | Curated sitemap for LLMs | Emerging proposal, not a ratified standard; low-cost to ship | | `llms-full.txt` | Full content dump for LLM ingestion | Variant of above; higher bandwidth, higher surface | ### 11. Validation Tools (reference only, no automatic calls) Paste the rendered JSON-LD into one of these public validators before publish. This skill does not hit them programmatically. | Tool | URL | Use for | |---|---|---| | Schema.org validator | `https://validator.schema.org/` | Strict schema.org conformance | | Google Rich Results Test | `https://search.google.com/test/rich-results` | Rich-result eligibility per Google's current rules | | Yandex structured data validator | `https://webmaster.yandex.com/tools/microtest/` | Cross-vendor sanity check | If a template in this skill ever conflicts with a validator result, trust the validator — schema.org and vendor rules evolve and this skill is a snapshot. ### 12. Workflow ``` Identify content type -> Pick primary schema from decision table -> Fill template from references/json-ld-schemas.md -> Add BreadcrumbList -> Wrap load-bearing passages in citation capsule markup -> Update llms.txt if site uses one -> Paste into validator -> Ship ``` ## Anti-Patterns - **Do NOT** copy a schema type used by a competitor without checking that your content actually fits it — mismatched types can trigger manual action penalties. - **Do NOT** leave `` in shipped JSON-LD. The schema is valid but the content is meaningless to retrieval systems. - **Do NOT** duplicate a primary schema type on the same page (two `Article` blocks). Answer engines pick one, often the wrong one. - **Do NOT** mark up invisible content. The schema must reflect what a human reader can see on the page. - **Do NOT** auto-fetch validator endpoints or vendor crawler docs from this skill. Templates are static; verification is a human step. - **Do NOT** assume `llms.txt` is a ratified standard in 2026. Ship it if your GEO budget allows, but do not treat it as table stakes. - **Do NOT** flip robots.txt from `Disallow` to `Allow` on AI crawlers without editorial + legal sign-off. It is a publishing-policy decision, not a technical one. ## Output Format When this skill produces a deliverable, the response should return one or more of: ``` SCHEMA BLOCK(S) --------------- File: [suggested filename, e.g., article.jsonld] Schema: [Article | OpinionArticle | FAQPage | HowTo | ...] Placeholders: [count of tokens to fill] ```json { ... } ``` CITATION CAPSULE(S) ------------------- Format: [HTML | Markdown] Count: [n] Placement: [first-500-words | per 800-word band] LLMS / ROBOTS SNIPPETS ---------------------- File: [llms.txt | robots.txt] Scope: [full site | section] VALIDATION CHECKLIST -------------------- - [ ] All replaced - [ ] Dates in ISO 8601 - [ ] Author Person linked to /about or author page - [ ] Pasted into schema.org validator and Google Rich Results Test ``` ## Quick Reference - **Default schema**: `Article` + `BreadcrumbList` + `Person` (author) - **Citation capsule rate**: 2-4 per 1500 words, first one in the first 500 words - **FAQPage rule**: Question text on page must match `Question.name` exactly - **Validator before publish**: schema.org + Google Rich Results Test, both - **llms.txt status (2026)**: emerging, not ratified — verify current --- ## References - See `${CLAUDE_SKILL_DIR}/references/json-ld-schemas.md` for full JSON-LD templates across 8+ schema types - See `${CLAUDE_SKILL_DIR}/references/llms-txt-template.md` for `llms.txt` template and robots.txt AI-crawler policy - See `../copywriting/references/aeo-geo-2026.md` for the AEO/GEO reference this skill implements