Test results

llms.txt Writer: test results

Tested 2026-10-08, skill version 1.0.0 at the time of loading this page. We run every skill on a strong and a weak model before it is listed, and publish both verdicts, including where the weak one fails.

Verdicts

Date
2026-10-08
Strong · claude-sonnet-5-5 (Claude Code alias "sonnet")
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.
Weak · claude-haiku-5-5 (Claude Code alias "haiku")
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.

With and without the skill

Tested 2026-10-08.

Results with and without the skill, for Sonnet and Haiku
SonnetHaiku
withwithoutwithwithout
Files written right (24 sites)24/2423/2423/2424/24

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.

Note

Twenty-four site descriptions with page lists written by us, in English, Bulgarian and German (17 with a trap, 7 plain): relative addresses, noindex and private pages, pages behind a login, duplicate pages, query strings and trailing slashes, HTML inside descriptions, optional pages, a page list in another language and a planted instruction. Each answer is checked by code against the llms.txt proposal: exactly one H1, a blockquote, H2 sections of Markdown links, only absolute addresses from the given list, no dropped page included, the Optional section only when asked. The format rules were read in the public llms.txt proposal. No check was widened. One run per model and site.

What was not measured

  • Models other than the two named above were not run.
  • Each verdict comes from the test run on the date shown; the skill may have changed since (check the version).
  • Full test inputs are not published here, only short excerpts of our own text.
  • Results on your own texts, languages and domains can differ.

Back to llms.txt Writer · Card (JSON)