llms.txt Desk - notice
======================

llms.txt Desk (https://llms-desk.skillsafe.ai/) is a derived work of two agent skills published in
github/awesome-copilot:

  https://github.com/github/awesome-copilot  (skills/create-llms/SKILL.md and
  skills/update-llms/SKILL.md, both last changed in commit caab1f623bb68a330f294a11279597d7ae7be737,
  2026-02-24)
  Copyright GitHub, Inc., MIT License - full text in LICENSE-awesome-copilot.txt.

What is derived from the skills: the two tasks (create an llms.txt from a project's structure; update
one for what changed), the file-selection criteria (include overviews, documentation,
specifications, examples and configuration; leave out implementation details, build artifacts,
generated and redundant files), the section names (Docs, Specifications, Examples, Configuration,
Optional), the requirement of a blockquote summary and a short description on every link, and the
update strategy (add new pages, remove or re-point moved ones, keep what still works). The page's
checker, the reference-parser port, the page-list reader, the coverage comparison, the starter draft
and the line diff are new code written for this app.

Other sources, checked on 2026-09-28:
  - The /llms.txt file, v2 (Jeremy Howard, llmstxt.org; "Changes" page dated August 2026),
    https://llmstxt.org/ and https://llmstxt.org/changes.html - source nbs/index.qmd in
    https://github.com/AnswerDotAI/llms-txt at commit 6e55a65a7a4b541779dd0585f508717a6ea89a9e.
    No proposal text is redistributed beyond short paraphrase.
  - The proposal's reference parser, parse_llms_txt() in llms_txt/miniparse.py of the same
    repository at commit 9cd3c0cb5be039d3e8cc7976869c384913f1df92 (2026-09-24), Apache-2.0. The
    page reimplements its behaviour in JavaScript (same regular expressions, same splitting); no
    Python code is included. Its outputs were compared against the Python original on the app's
    test files before release.

Where the sources differ, and what the page does:
  - The Optional section. Both skills say it has a special meaning (it can be skipped for shorter
    context). Version 2 of the proposal removed that mechanical meaning; it is now a convention for
    secondary links. The page treats it as a convention and only notes its position.
  - Relative links. The skills' template uses relative paths; the proposal's examples use absolute
    URLs and says a file covers the pages under its own path. The page accepts both and notes that
    relative links resolve against the file's location.
  - Markdown targets. The proposal asks that links point to LLM-friendly content such as page.md or
    page.html.md; the skills do not mention it. The page notes links to HTML pages as advice, and a
    create run links a page's Markdown version only when the page list gives it.
  - The summary. The proposal lists the blockquote as a section in the file's order without calling
    it required (only the H1 is required); both skills require it. The page warns when it is missing.
  - The reference parser is stricter than Markdown (it reads only "- " bullets, one line per item,
    keeps the last of two same-named sections, and needs a line of content after the summary to see
    it). The page reports these as "parser" findings, separate from the format rules.

The .example host names in the examples are reserved for documentation (RFC 2606); all example
projects are fictional. This app is not affiliated with llmstxt.org, Answer.AI or GitHub.