llm-friendly-site-architect

LLM Friendly Site Architect

Use this skill after semantic core, URL architecture, internal link graph, and technical SEO/schema foundations exist.

Read references/llm-readable-architecture.md before producing a full LLM discovery plan.

Owns

• llms.txt structure;
• LLM discovery plan;
• visible answer blocks and concise summaries;
• source trails;
• entity pages and topic maps for agents;
• markdown alternate policy;
• AI crawler policy handoff;
• no-duplicate-content guardrails for LLM-readable surfaces.

Does Not Own

• Search Console or assistant citation monitoring execution;
• technical schema implementation;
• URL/canonical decisions;
• hidden content;
• external authority placement.

Workflow

1. Read url-map.yaml, internal-link-graph.yaml, schema/head policy, and visible content model.
2. Identify canonical public pages that should be discoverable by assistants.
3. Draft or audit llms.txt sections with only public useful URLs.
4. Define visible direct-answer, TLDR, facts, source, or FAQ blocks only where page role supports them.
5. Define source-trail and citation-ready evidence surfaces.
6. Define markdown alternate rules for longform/docs pages when useful, with noindex or alternate policy as appropriate.
7. Define entity pages and topic pages that clarify who, what, and how the site works.
8. Record crawler policy items that need current primary-source verification.
9. Produce an LLM discovery report using assets/llm-discovery-report.template.md.

Non-Negotiables

• Human-visible content first.
• No hidden bot-only facts.
• No doorway pages.
• No FAQPage schema unless FAQ is visible.
• No automatic TLDR on every short news page by default.
• No extra public story bodies beyond the declared content model.
• llms.txt is a helpful convention, not a guaranteed ranking or citation factor.
• Crawler names, policies, and IPs require primary-source verification before production changes.