Generate
Give the tool a URL, a name, a one-line summary and any sections you want listed. It writes the file in the shape described by the llmstxt.org convention, with every link made absolute.
LLMS.TXT · GENERATE & CHECK
Write an llms.txt
your site can be found by.
llms.txt is a plain-text file that tells language models what your site is and which pages matter. Fill in the form and get a well-formed file you can drop at the root of your domain — or paste an existing file and get a line-by-line report of what is wrong with it.
Everything runs in this browser tab. Nothing you type is sent to MySkills.
THE TOOL
Generate a file, or check the one you have.
The base address of your site. Relative paths such as /docs/intro are turned into absolute links against it.
Used as the # Title line. Left empty, the host name is used instead.
This becomes the > summary blockquote directly under the title. One sentence, plain wording.
Free text after the summary. Useful when the summary alone does not say enough.
Each section becomes a ## Heading followed by markdown links. One link per line, in the form Label | /path | optional description. A bare /path works too — the label is then taken from the last segment.
No sections yet. You can still publish a valid file with a title and a summary, but a file with sections is more useful to a model.
Nothing has left your browser. Fill in the fields and generate a file.
OUTPUT
Your llms.txt
Paste the text or the raw file contents. The report names the exact line for every finding.
Nothing has left your browser. Paste a file to check it.
Check
Paste a file you already have. The checker reads it line by line and reports each problem with the line number and a concrete fix. It reports findings, not a score.
Publish it yourself
Copy the result to a file named llms.txt and put it at the root of your domain, for example https://example.com/llms.txt. No hosting service needed.
WHAT LLMS.TXT IS
A suggestion to a model, not a command.
llms.txt is a voluntary convention proposed by Jeremy Howard and friends at llmstxt.org. The idea is simple: if a website publishes a markdown file at /llms.txt that says what the site is and which pages are worth reading, a language model that looks for it gets a much better starting point than a raw HTML scrape.
The convention itself is short. A file starts with an H1 title, then a blockquote with a one-line summary, then optional prose, then any number of ## sections, each holding markdown links to the pages that matter. That is the whole format. This page writes that format and checks that your file actually follows it.
What it is not
It is not a robots file, and it does not replace one. It is not a sitemap, it does not submit your site anywhere, and it does not set crawl rules. It does not tell a search engine to rank you higher, and it does not make a model cite you. Adding an llms.txt file guarantees nothing: whether any given model reads it, and whether that changes an answer, is up to that vendor.
What it plausibly does is make your site easier to summarise correctly if a model does decide to look. If you do it, keep it honest. A short file that describes what you actually publish beats a long one padded with every URL you have.
LIMITS
What this tool does not do.
- It does not fetch your site. There is no request to your URL, so the tool cannot read your navigation, guess your pages or find broken links. You list the pages; it formats them.
- It does not submit anything. No search engine, no model provider and no directory is contacted. You publish the file yourself.
- It does not score your site. The checker reports concrete facts — a missing title on line 1, a relative link on line 14 — and stops there. There is no invented grade, because a number that no one can explain is just marketing.
- It does not verify your links resolve. A well-formed absolute URL can still point at a 404. Check them in your own logs.
- It follows the published convention, not a vendor dialect. If a specific tool expects a slightly different shape, that tool's documentation wins.
FAQ
Before you publish the file
Does MySkills send my site URL or my content to a server?
No. The generator and the checker are plain JavaScript in your browser tab. Your typing is never sent anywhere, and there is no request to the site URL you type in. The page loads only its own stylesheet, its own script and this site's optional analytics script.
Where does the llms.txt file go?
At the root of your domain, named llms.txt. If your site is https://example.com, the file belongs at https://example.com/llms.txt — same folder as robots.txt. The checker will flag relative links inside the file, because a model reading the file may not know what the pages are relative to.
Will adding an llms.txt file make ChatGPT or Gemini cite me?
No guarantee, and any tool that promises one is selling something. The convention is new and adoption is voluntary. Publishing an accurate file is cheap and cannot hurt; treating it as a ranking tactic is a mistake.
How long should the file be?
Short enough to be read. Most useful examples are under a page: a title, a summary, and the handful of pages that represent the site. A file that lists every URL is the same as having no summary.
Can I check a file that is already live?
Open https://example.com/llms.txt in your browser, select the text and paste it into the Check tab. The report gives the exact line for each problem, so you can fix them one at a time.
Does the checker replace my own review?
No. It checks the shape of the file against the convention. It cannot judge whether the summary is accurate, whether the pages are the ones worth pointing at, or whether the links still work.