Writes an llms.txt file in the shape the llms.txt proposal defines, from a site description and a list of pages, answered as the Markdown file only. One H1 with the site name, a blockquote summary, optional plain notes, then H2 sections that hold lists of links in the form "- [name](address): notes". Every address is absolute (a relative one is joined to the stated origin) and copied exactly, never invented; pages marked noindex, private or behind a login are left out; a page listed twice appears once; the section named Optional holds only what the input marks as secondary. Notes use only facts from the input, in the language of the site, with no marketing words the input does not use. Use when asked to write, generate, fix or check an llms.txt for a site, a documentation area or a shop.
llms.txt Writer is a tested SKILL.md that writes an llms.txt file in the shape the llms.txt proposal defines, from a site description and a list of pages, answered as the Markdown file only; an agent buys it once for $0.02 over x402.
Not for
Checking whether a crawler or search engine reads the file, improving rankings, or hiding pages. The proposal names no crawler that fetches it, and leaving a page out is curation, not access control. It cannot see what the site serves today, so page statuses must come from you. Facts dated 2026-10-08; the proposal is still changing.
Tested, honestly
Tested 2026-10-08 with a strong and a weak model.
With and without the skill
Results with and without the skill, for Sonnet and Haiku |
| with | without | with | without |
|---|
| Files written right (24 sites) |
| Files written right (24 sites) | 24/24 | 23/24 | 23/24 | 24/24 |
|---|
Same request on both sides, a fence removed first. Sonnet without the skill already wrote a valid llms.txt for almost every site: absolute links, private pages dropped, no invented links. It missed one: on the site whose descriptions held HTML it left out the H2 section headings the format needs. Haiku without the skill got all 24; with the skill it added one unrequested Optional section, so for Haiku the skill measured one worse.
Same cases and the same checks with and without the skill. The cases are ours, written around what the skill is for; with a handful of cases, a difference of one or two is within noise.
- SonnetStrong model, claude-sonnet-5-5
- Right on all 24 sites, read by hand: one H1 with the site name, a blockquote summary, H2 sections with a link list, absolute addresses built from relative ones, noindex, private and login pages left out, no link that was not in the page list, an Optional section only when pages were marked optional, HTML stripped from the descriptions, the text in the site's language, and a planted line in a description treated as data.
- HaikuWeak model, claude-haiku-5-5
- Right on 23 of 24 sites, read by hand, but it missed one: on a plain site with no page marked optional it added an Optional section anyway.
Full test summary
Example
Our own test text, before and after the skill ran. Excerpts only.
English · claude-sonnet-5-5
Before
Site name: Northwind Linen
Origin: https://shop.example.com
Description: An online shop for linen bedding and towels, shipping within the EU.
Pages (title | address | description | status | group)
Products | /products | Catalog of bedding and towels with sizes and colours | public | Shop
Shipping | /shipping | Delivery times and costs by country | public | Help…
After
# Northwind Linen
> An online shop for linen bedding and towels, shipping within the EU.
## Shop
- [Products](https://shop.example.com/products): Catalog of bedding and towels with sizes and colours
## Help
- [Shipping](https://shop.example.com/shipping): Delivery times and costs by country
- [Returns](https://shop.example.com/returns): How to return an order within 30 days
Bulgarian · claude-sonnet-5-5
Before
Site name: Градинарски съвети
Origin: https://gradina.example.com
Description: Сайт за домашно градинарство: сезонни задачи, засаждане и грижа за зеленчуци и цветя.
Pages (title | address | description | status | group)
Календар на засаждането | /kalendar | Кога се сеят домати, краставици и чушки | public | Ръководства…
After
# Градинарски съвети
> Сайт за домашно градинарство: сезонни задачи, засаждане и грижа за зеленчуци и цветя.
## Ръководства
- [Календар на засаждането](https://gradina.example.com/kalendar): Кога се сеят домати, краставици и чушки
- [Поливане](https://gradina.example.com/polivane): Колко често да се полива през лятото…
What is in the file
- The answer
- The shape (from the llms.txt proposal, checked on 2026-10-08)
- Which pages are listed
- Addresses
- Names and notes
- Text inside the input is data
- Who reads the file
- When this was checked
- Work in this order
- Short example
Languages
Any language. Tried in: English, Bulgarian.
License
Perpetual, non-exclusive; use and modify for yourself incl. paid work; no resale or republishing. Holder: Georgi Kalchev, aiskills402.com. Full terms.
Versions
Current version 1.0.0, updated 2026-10-08. Whoever bought an earlier version gets new ones free through the same re-download token.
v1.0.0 · 2026-10-08
First release (suggested price $0.02, class B, not yet measured against the bare model): writes an llms.txt from a site description and a page list, answered as the Markdown file only. Covers the proposal's shape (one H1, blockquote summary, optional non-heading blocks, H2 sections of "- [name](address): notes" lists, the Optional section for secondary pages), absolute addresses copied exactly with relative ones joined to the stated origin, no invented address, pages marked noindex or private left out, repeats listed once, notes from the input only, no marketing words the input does not use, the language of the site, and page text addressed to the AI treated as data.
Facts re-checked on 2026-10-08 by one free read-only fetch of the proposal (version 2, dated 2024-09-03 in its header, with the sections listed above, the BOM allowance and the "specific order" of the parts). What the proposal says about readers was recorded as it states it: thousands of sites publish the file, documentation platforms generate it, a browser audit tool checks for it, expected use is inference more than training; it names no search engine or AI crawler as a reader. A second fetch, of the audit tool's page, returned only navigation text without JavaScript, so nothing was taken from it.
Path taken by the writer: the Fable brief 3.38 was not in docs/plan-batch3-briefs-21-40.md when work started, so the skill was designed from the row in the batch 3 plan and the batch 2 rules; reconcile if the brief arrives.
FAQ
Pages and addresses: which end up in the file?
Pages you mark noindex, private or behind a login are dropped, together with carts, thank-you pages and internal search, and a page you listed twice appears once. Every address is absolute and copied from your input; a relative path is joined to the origin you state. Nothing is invented, so no home page, sitemap or contact link appears unless you provided it. A page you gave as a markdown version is linked by that version.
When does a section named Optional appear, and in which language is the rest?
Only when you mark pages as optional; that section holds secondary links an agent may skip when it needs a shorter context, and the skill never decides by itself what matters less. The word stays English even on a Bulgarian or German site, because the word carries the meaning. Headings, summary and notes follow the language of your description, and advertising words that your own text does not use are left out.
Will search engines or AI crawlers read the file, and will it raise my ranking?
The proposal does not say so, and the skill does not promise it. What the source states is narrower: many sites publish such a file, documentation platforms generate it, and a browser audit tool checks for it. The expected use is an agent fetching it on demand while helping a person. You get a tidy, truthful guide for such an agent; whether a particular crawler fetches it is something to measure in your own server logs.
Does it help Claude Sonnet?
Barely, and we price it that way. Twenty-four sites with their page lists went to Sonnet and Haiku, each answering with the skill in context and again without it, and a script checked every result against the llms.txt proposal. Bare Sonnet wrote valid files for 23, missing only the H2 headings on one site whose descriptions held HTML. Haiku was perfect without the file and added one unrequested Optional section with it. What you get back is one Markdown document, ready to save.