diff --git a/README.md b/README.md index 81d43da..dc6707b 100644 --- a/README.md +++ b/README.md @@ -14,13 +14,13 @@ See [AGENTS.md](./AGENTS.md) for guidelines on creating and submitting skills. ## Skills -| Skill | Description | -|-------|-------------| -| [aws-bedrock-evals](./skills/aws-bedrock-evals/) | Build LLM-as-judge evaluation pipelines with Amazon Bedrock Evaluation Jobs | -| [react-effect-guide](./skills/react-effect-guide/) | Identify unnecessary useEffect usage in React code and recommend the correct alternative pattern | -| [aws-well-architected-review](./skills/aws-well-architected-review/) | Enforces AWS Well-Architected best practices during infrastructure generation and reviews existing infra code for anti-patterns with actionable fixes | -| [agent-web-compatibility](./agent-web-compatibility/) | Audits and redesigns websites for dual consumption optimised | -| [figma-design-audit](./skills/figma-design-audit/) | Audits a Figma design against its design system. Checks color variables, font styles, numeric variables (padding/corner radius/gap), and unnecessary nesting. Posts inline comments on the Figma canvas for every violation found. | +| Skill | Description | +| -------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | +| [aws-bedrock-evals](./skills/aws-bedrock-evals/) | Build LLM-as-judge evaluation pipelines with Amazon Bedrock Evaluation Jobs | +| [react-effect-guide](./skills/react-effect-guide/) | Identify unnecessary useEffect usage in React code and recommend the correct alternative pattern | +| [aws-well-architected-review](./skills/aws-well-architected-review/) | Enforces AWS Well-Architected best practices during infrastructure generation and reviews existing infra code for anti-patterns with actionable fixes | +| [agent-web-compatibility](./skills/agent-web-compatibility/) | Audits websites for AI consumption — agentic task completion, AEO, and GEO. AI-optimized, human-compatible. | +| [figma-design-audit](./skills/figma-design-audit/) | Audits a Figma design against its design system. Checks color variables, font styles, numeric variables (padding/corner radius/gap), and unnecessary nesting. Posts inline comments on the Figma canvas for every violation found. | ## License diff --git a/skills/agent-web-compatibility/README.md b/skills/agent-web-compatibility/README.md index 772d01e..e836f70 100644 --- a/skills/agent-web-compatibility/README.md +++ b/skills/agent-web-compatibility/README.md @@ -26,7 +26,7 @@ This skill is used by **practitioners building or redesigning websites for busin | Audience | Use Case | | ------------------------ | ------------------------------------------------------- | | Web agencies | Differentiate packages with "agent-ready" websites | -| No-code platform teams | Bake dual consumption into templates at scale | +| No-code platform teams | Bake AI compatibility into templates at scale | | Booking software vendors | Make widgets agent-completable by default | | GEO / SEO consultants | Extend beyond citation into transactional AI visibility | @@ -43,13 +43,7 @@ This skill is used by **practitioners building or redesigning websites for busin ## Installation ```bash -npx skills add antstackio/skills -``` - -Or install directly: - -```bash -npx skills add /skills +npx skills add antstackio/skills --skill agent-web-compatibility ``` --- @@ -125,6 +119,8 @@ Every audit produces five deliverables: Detailed implementation specs with ready-to-paste JSON-LD examples are included for: +**Transactional — agentic booking** + | Vertical | Schema Type | Key Fields | | -------------------- | ---------------------------- | ----------------------------------------------------------- | | Healthcare / Clinics | `MedicalClinic`, `Physician` | `medicalSpecialty`, `hasCredential`, `availableService` | @@ -132,6 +128,24 @@ Detailed implementation specs with ready-to-paste JSON-LD examples are included | Salons / Spas | `BeautySalon`, `DaySpa` | `makesOffer` with duration, `employee` specialisations | | Local E-commerce | `OnlineStore`, `Product` | `deliveryLeadTime`, `areaServed`, `hasMerchantReturnPolicy` | +**Informational — AEO + GEO** + +| Vertical | Schema Type | Key Fields | +| ------------------------------ | ---------------------------------------- | ------------------------------------------------------- | +| Blog / Editorial | `BlogPosting`, `Article` | `author` with `hasCredential`, `dateModified` | +| Company Marketing | `Organization`, `WebSite` | `hasOfferCatalog`, `areaServed`, `sameAs` | +| Personal Portfolio | `Person`, `ProfilePage` | `knowsAbout`, `hasOfferCatalog`, `sameAs` | +| Documentation / Knowledge Base | `TechArticle`, `HowTo` | `HowToStep`, `proficiencyLevel`, `dateModified` | + +**Navigational — discovery + contact** + +| Vertical | Schema Type | Key Fields | +| --------------------- | ------------------------------------------- | ------------------------------------------------------- | +| Agency / Consultancy | `ProfessionalService` | `serviceType`, `areaServed`, `employee` with credential | +| SaaS / Web App | `SoftwareApplication` | `featureList`, `applicationCategory`, `offers` | +| Education / Courses | `Course`, `EducationalOrganization` | `instructor`, `hasCourseInstance`, `educationalCredentialAwarded` | +| Travel / Hospitality | `LodgingBusiness`, `TravelAgency` | `checkinTime`, `amenityFeature`, `petsAllowed` | + --- ## Why Not Just Rely on Zomato / Practo / Google? @@ -166,7 +180,7 @@ agent-web-compatibility/ ### 1. Install ```bash -npx skills add antstackio/skills +npx skills add antstackio/skills --skill agent-web-compatibility ``` ### 2. Trigger with a test prompt diff --git a/skills/agent-web-compatibility/SKILL.md b/skills/agent-web-compatibility/SKILL.md index beb2072..a70e4a5 100644 --- a/skills/agent-web-compatibility/SKILL.md +++ b/skills/agent-web-compatibility/SKILL.md @@ -1,6 +1,6 @@ --- name: agent-web-compatibility -description: Audit and redesign websites for dual consumption — optimised for both human visitors and AI agents acting on their behalf. Use this skill when someone wants to make a website agent-compatible, improve how AI assistants discover or recommend their site, build trust signals for AI-mediated transactions, or ensure their booking/reservation flow works for AI agents. Triggers on phrases like "agent-friendly website", "AI can't book on my site", "make my site work with AI assistants", "optimise for AI agents", "dual consumption", "agentic web", "my site isn't getting picked by AI", or any request to audit or redesign a website for AI compatibility. Also triggers when building new websites for clinics, restaurants, salons, local e-commerce, or event booking where agentic discovery and completion is a goal. +description: "Audit and redesign websites for AI consumption — covering agentic task completion (can an AI book/order on a user's behalf?), AEO (Answer Engine Optimization — can content surface as a direct answer in voice or featured snippets?), and GEO (Generative Engine Optimization — will AI systems like ChatGPT, Perplexity, and Google AI Overviews cite this content?). AI-optimized, human-compatible: the goal is AI performance; human experience is a constraint, not a deliverable. Use this skill when someone wants to make a website agent-compatible, improve how AI assistants discover or recommend their site, build trust signals for AI-mediated transactions, ensure their booking/reservation flow works for AI agents, or get their content cited in AI-generated answers. Triggers on phrases like \"agent-friendly website\", \"AI can't book on my site\", \"make my site work with AI assistants\", \"optimise for AI agents\", \"agentic web\", \"my site isn't getting picked by AI\", \"AEO\", \"GEO\", \"AI citations\", \"featured snippets\", \"AI Overviews\", or any request to audit or redesign a website for AI compatibility. Also triggers when building new websites for clinics, restaurants, salons, local e-commerce, or event booking where agentic discovery and completion is a goal." license: MIT metadata: author: antstackio @@ -9,20 +9,26 @@ metadata: # Agent Web Compatibility -Audit and redesign websites so they perform well for **both** human visitors and AI agents acting on a user's behalf. +Audit and redesign websites for AI consumption — so agents can find, trust, and transact on them. The goal is AI performance; human experience is a constraint: optimizations must not break what already works for human visitors, but improving human UX is out of scope. -When an agent is asked _"find me a dermatologist near Indiranagar available this Thursday"_ it doesn't click buttons or read hero images — it parses structured data, verifies trust signals, and picks the option it can confidently complete. This skill addresses all three stages of that decision. +Three AI consumption modes are in scope: + +- **Agentic task completion** — when an agent is asked _"find me a dermatologist near Indiranagar available this Thursday"_ it parses structured data, verifies trust signals, and completes the booking without human intervention. +- **AEO (Answer Engine Optimization)** — when someone asks a voice assistant or search engine a question, your content surfaces as the direct answer. +- **GEO (Generative Engine Optimization)** — when ChatGPT, Perplexity, Google AI Overviews, Claude, or Gemini generate an answer, your content gets cited or recommended. **Best fit:** independent clinics, restaurants, salons, local e-commerce, event venues. Not for chains already integrated into aggregator APIs at scale. --- -## The Three-Layer Framework +## The Five-Layer Framework ``` 1. DISCOVERABILITY — Can the agent find and understand you? 2. PREFERABILITY — Does the agent trust you enough to recommend you? 3. COMPLETABILITY — Can the agent finish the transaction without breaking? +4. ANSWERABILITY — Does your content surface as a direct answer (AEO)? +5. CITABILITY — Will AI systems cite or recommend your content (GEO)? ``` Work through layers in order. Read the vertical reference file before auditing. @@ -69,6 +75,75 @@ Agents evaluate verifiable signals, not marketing copy. Every vertical has speci --- +## Layer 4: Answerability (AEO) + +AEO targets voice assistants, featured snippets, and "People Also Ask" boxes. The goal: your page is the source of the direct answer, not just a result in a list. + +**Answer-format content** — place a concise direct answer (2–4 sentences or a tight list) at the top of any page targeting a question query. Do not bury the answer after marketing copy. + +**`FAQPage` schema** — mark up Q&A content on service pages, about pages, and dedicated FAQ pages. Every `Question` needs an `acceptedAnswer` with `text` in plain prose (not HTML). + +```json +{ + "@type": "FAQPage", + "mainEntity": [{ + "@type": "Question", + "name": "Does the clinic offer teleconsultation?", + "acceptedAnswer": { + "@type": "Answer", + "text": "Yes. Teleconsultation is available Monday to Saturday, 9 AM to 8 PM. Book via the website or call +91-80-XXXXXXXX." + } + }] +} +``` + +**`HowTo` schema** — use for any step-by-step content (how to book, how to prepare for a procedure, how to place an order). Each `HowToStep` needs `name` and `text`. + +**`Speakable` schema** — mark sections suitable for voice readout with `speakable` on the page entity. Target the business description, hours, and key service summary. + +**Page structure for snippet capture:** +- H1 = the question as asked ("How do I book a same-day appointment at [Clinic]?") +- First paragraph = direct answer, ≤ 60 words +- Use `