Make a website readable by AI
Help agents retrieve a clean copy of a public page without breaking the version people use in a browser.
Make a website readable by AI when the browser shows a complete article but a text client only gets the navigation. The llms-visibility skill gives a coding agent checks for the first response from the server, the main content and an equivalent Markdown representation.
A documentation site is a good place to try it. A code example or pricing table should say the same thing in either format. The guide covers how a client discovers the alternative, how it requests one format and how a cache keeps the two responses apart. It also separates measuring client traffic from the work of serving pages.
What it does
- Find articles whose main text appears only after JavaScript runs.
- Keep the heading hierarchy, links, tables and code examples meaningful.
- Expose a Markdown path for each type of public content page.
- Check the response when a client accepts either HTML or Markdown.
- Decide which pages belong in a short index for agents.
Install and use
- Give your coding agent access to an authorised website project and its serving configuration. Use curl locally and an internet connection for public checks.
- Ask it to make one help article available in both formats, then inspect the response headers and bodies.
- Compare the article in a browser and a text client before applying the change to other templates.
FAQ
Source
Made by Evil Martians. Original skill: llms-visibility, published on GitHub under the MIT licence.
View on GitHub