| name | LLMs.txt Format |
| description | Use this skill when the AI SEO agent needs to generate an llms.txt entry for a spec page, or when the deploy agent needs to update the site-wide llms.txt file. Documents the llms.txt standard, the entry format for robertrevans.com spec pages, and how individual page entries compose into the site-level index. |
| version | 0.1.0 |
LLMs.txt Format for Spec Pages
What llms.txt is
llms.txt is a site-level file that tells AI systems what content is on the site, where it lives, and what each page covers — in a format optimized for AI consumption rather than human reading or search crawler ranking.
The analogy is robots.txt: a simple, machine-readable file at the root of a site that gives AI systems structured direction rather than making them crawl and infer everything. Unlike robots.txt (which restricts access), llms.txt invites AI systems in and tells them exactly what they'll find.
The standard is emerging (site: llmstxt.org). This skill defines how robertrevans.com implements it for spec pages.
Site-level file
The site-wide llms.txt lives at:
https://robertrevans.com/llms.txt
Local path:
[robertrevans.com project]/llms.txt
Every time a new spec page is deployed, the deploy agent appends a new entry to this file. The file grows monotonically — entries are never removed, only added.
Entry format
The llms.txt standard uses Markdown, not a custom key-value format. Each spec page produces one bullet-point entry. The format:
- [Concept Name](https://robertrevans.com/[category]/[page-name]/): [2-3 sentence factual description. What it is, how it works, who built it, what problem it solves.]
Rules:
- Description must be factual and specific. No marketing language.
- 2-3 sentences max. Lead with what it is, follow with how it works.
- Weave in the key questions the page answers — AI systems use this to decide whether the page is relevant to a query.
- All content required. No "TBD."
Example entry
- [Spec Page Agentic Team](https://robertrevans.com/specs/spec-page-agentic-team/): A Claude Code plugin that takes any idea, brief, or URL and produces a complete, deployed HTML spec page using a seven-agent pipeline — discovery, blueprint, build, code-review, SEO, AI SEO, and deploy. Built by Robert Evans. Covers: what the pipeline does, how each agent works, how to use the plugin, and how AI SEO differs from traditional SEO.
Site-level file structure
The full llms.txt file uses standard Markdown with H2 sections per category. When creating the file for the first time:
# Robert Evans — Specs, Tools, and Concepts
> CTO and systems builder publishing spec pages, tools, and architectural concepts at robertrevans.com. Each page is built on The Point CSS framework and optimized for both traditional SEO and AI discoverability.
## Specs
## Tools
## Concepts
H2 sections are created once and never removed. Entries accumulate under the right section monotonically — never edited, never deleted.
Where the AI SEO agent produces the entry
The AI SEO agent writes the entry artifact to:
docs/briefs/NNN-[concept-name]/NNN.07-llmstxt-[concept-name].txt
This file contains ONLY the entry block for this page (no header, no other entries). The deploy agent reads this file and appends it to the site-wide llms.txt.
What the deploy agent does with it
On deploy, after copying the HTML file, the deploy agent:
- Reads
NNN.07-llmstxt-[concept-name].txt (the bullet-point entry)
- Determines the category from the brief:
specs, tools, or concepts
- Finds the matching H2 section in
robertrevans.com/llms.txt (## Specs, ## Tools, or ## Concepts)
- Appends the bullet-point entry as a new line under that section
- If the section doesn't exist, creates it before appending
Never edits or removes existing entries. The file grows monotonically.
Note on adoption: As of mid-2026, no major LLM provider has confirmed crawling llms.txt files. The ROI is speculative. We generate it because the cost is low and it future-proofs the site — but this is not a primary optimization signal. Treat it as infrastructure, not strategy.
Relationship to AP annotations
The topics: field in the llms.txt entry should match the topic attribute values used in the AP reading layer annotations on the same page. They are two representations of the same information:
- AP annotations give AI agents precise XPath addresses within the page
- llms.txt gives AI systems a pre-extracted summary without needing to fetch the page at all
Together they mean: an AI system can answer questions about this page without fetching it (via llms.txt), and if it does fetch the page, it can navigate directly to any specific answer (via AP annotations).
