SEO/AEO Master Playbook

---
name: seo-aeo-master-playbook
description: >
  The one prompt to rule them all — make any website maximally discoverable by every
  search engine, every AI assistant, and every autonomous agent on the open web.
  Classic SEO + AEO (Answer Engine Optimization) + citation infrastructure in a single
  idempotent audit-and-ship brief. Covers llms.txt, MCP servers, JSON-LD, AI crawler
  allowlists, segmented sitemaps, IndexNow, per-route OG images, and Core Web Vitals.
  Paste into Claude Code, Cursor, or Windsurf from your repo root and walk away.
license: CC BY 4.0
---

# SEO/AEO Master Playbook — The One Prompt to Rule Them All

This skill is a copy-paste master prompt. Drop it into any capable coding agent
(Claude Code, Cursor, Windsurf, Aider) from inside your repo's root directory,
replace the two `<placeholders>` on the first line, and walk away for an hour.

It's **idempotent** — run it again after changes and it will audit and fill gaps
without duplicating work.

## Why this exists

In 2026, making a site discoverable is a three-job brief, not one:

1. **Be crawlable** — classic SEO. Sitemaps, robots, fast pages, clean URLs.
2. **Be ingestible** — AEO (Answer Engine Optimization). llms.txt, MCP server,
   machine-readable API, OpenAPI spec, Markdown export, AI crawler allowlist.
3. **Be citable** — structured data so answer engines can quote you *with a URL
   attached*. JSON-LD on every page.

Missing any one of the 14 pieces below creates a dead spot where a specific class
of agent can't find, can't parse, or can't cite you. This prompt covers all 14.

## How to use

1. Open Claude Code (or Cursor / Windsurf) at the root of your repo.
2. Paste the prompt block below.
3. Replace `<YOURSITE>` with your site name and `<ONE-LINE DESCRIPTION>` with
   what the site is.
4. Let it run the audit pass first, then the ship pass.
5. Read the status file it writes at the end. Do the registration steps
   yourself (Search Console, Bing Webmaster, IndexNow key, smithery.ai).

## The prompt

You are helping me make this site maximally discoverable by every search
engine, every AI assistant, and every autonomous agent on the open web. The
site is <YOURSITE> (canonical URL: <https://yoursite.com>), and it is a
<ONE-LINE DESCRIPTION>.

This is a three-job brief: classic SEO (be crawlable), AEO (be ingestible),
and citation infrastructure (be citable). Do all three. Do not skip any of
the 14 pieces below unless I tell you to, and when you skip one, tell me WHY
in the final report.

## Audit first (read-only pass)
Before writing any code, inventory what already exists:
1. robots.txt / robots route
2. sitemap.xml / sitemap route
3. llms.txt and llms-full.txt at the site root
4. .well-known/ai-plugin.json and .well-known/openapi.yaml
5. Any MCP server script in the repo
6. JSON-LD structured-data helpers
7. Per-page metadata exports
8. Per-route Open Graph image generators
9. IndexNow submission script or library
10. RSS / feed route
11. oEmbed route
12. Any nginx / Caddy / Cloudflare config snapshots
13. Framework config (rendering mode, image config)
14. Topic hub and "best of" listicle routes

Report what is present, what is stale, and what is missing. Do not change
anything yet.

## Then ship the missing pieces

For each of the 14 below that is missing or incomplete, build it. Favor the
smallest working version first; we can extend later. Follow the conventions
already in the repo (framework, TS strictness, path aliases, lint rules).

1. llms.txt — minimum viable manifest at the site root.
2. llms-full.txt — a build-time script that dumps every public page plus catalog item in plain text. Target size: as large as practical. Declare CC BY 4.0 at the top.
3. .well-known/ai-plugin.json — a ~15-line manifest pointing at the OpenAPI spec.
4. .well-known/openapi.yaml — real contract for every public endpoint.
5. MCP server — a small Node or Python server exposing 4–8 tools that proxy the public API. Do NOT hit the database directly. Ship a README snippet telling users how to install it in Claude Desktop's config.
6. Public API — JSON default, Markdown via ?format=md or Accept: text/markdown. Markdown must include a citation footer with the canonical URL and license. At minimum: list, detail, categories.
7. robots — explicit allowlist for every named AI crawler (GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, anthropic-ai, Claude-Web, PerplexityBot, Perplexity-User, Googlebot-Extended, Applebot-Extended, CCBot, cohere-ai, Diffbot, FacebookBot, Meta-ExternalAgent, MistralAI-User, YouBot, DuckAssistBot, Amazonbot, Bytespider). Disallow auth / admin / private. Link every sitemap explicitly.
8. Segmented sitemaps — one per surface, not one monolith. Scale priority by live signals (install count, freshness, verification). Never ship a single sitemap over 1k URLs.
9. Topic hubs and "best of" listicles — at least 10 of each. Target real search intent. Every list page emits an ItemList JSON-LD. Every one is in the sitemap at priority 0.85+.
10. JSON-LD helpers — one helper module exporting: SiteJsonLd, BreadcrumbJsonLd, ItemListJsonLd, FAQJsonLd, HowToJsonLd, SoftwareApplicationJsonLd (or ProductJsonLd), ArticleJsonLd, BlogPostingJsonLd. Wire them into every page that matches the shape. SiteJsonLd goes in the root layout and MUST emit Organization + WebSite + SearchAction.
11. Per-route OG images — one generator on every dynamic detail route (items, blog posts, creators, lists).
12. Per-page metadata — title template in the root layout, plus per-page title / description / canonical / Twitter card on every page with dynamic content.
13. IndexNow — a small library that POSTs to api.indexnow.org, called from every publish / update mutation. Never throw. Ship the public verification key file.
14. Perf pass — measure LCP / INP / CLS on the top 5 routes. If LCP > 2.5s, fix it. If the site is behind nginx, add an HTML cache layer for public SSR routes that bypasses on session cookie. If behind Cloudflare, configure cache rules and bot management to allow the AI allowlist.

## Final report format

When done, write a short status file with:
- Baseline (what was there)
- Shipped (what you built)
- Skipped (with reasons)
- Next actions for me (registration: Google Search Console, Bing Webmaster, IndexNow key, submitting sitemaps, submitting the MCP server to smithery.ai)
- Measurement plan (which queries to check in Search Console weekly, which dashboards to watch, what "success" looks like in 30/60/90 days)

## Hard rules
- Never break existing routes. If you touch metadata, verify the page still renders.
- Never include PII or private routes in llms.txt / sitemaps / JSON-LD.
- Never keyword-stuff. If a human reader can tell it was written for SEO, rewrite.
- Every cite-able page must include the license and the canonical URL in its Markdown export.
- If you hit a blocker, write it in the status file and continue with the next piece. Do not stall the whole pass on one issue.

---

License: CC BY 4.0 — copy it, paste it, fork it, ship it. Attribution appreciated:
https://a-gnt.com/agents/skill-seo-aeo-master-playbook