Claude CodePartly tested🌏 World

Three prompts to have Claude Code build a LaTeX tool that lives entirely in the URL #fragment

Inspired by Tiny Little Office, Simon Willison used three Claude Code prompts to build a single-page LaTeX renderer whose (optionally compressed) content lives in the URL #fragment, so the link is the document; follow-ups added an iframe sandbox against XSS and URL/Gist loading.

Visit site Claude Code ↗
Pricing
Paid
Region
Global
Where it's from
Overseas
Added
Oct 17, 2026
Tested on
Oct 17, 2026
Version we used
tools.simonwillison.net/latex (checked 2026-10-17); prompts from simonwillison.net 2026-10-15
Categories
CodingProductivity
Original author
Simon Willison @simonw ↗
Our test
Partly tested
Cost / time
Needs Claude Code (paid). The output is a static single page you can host free on GitHub Pages or any static host.
Source
Simon Willison's Weblog ↗tools.simonwillison.net ↗

The task

Build a tiny tool with no backend or database where whatever you write becomes a bookmarkable, shareable link.

Steps

  1. Open Claude Code in your tools repo and send the first prompt (below): LaTeX in the #fragment, an optional prefix for compression with a LaTeX-tuned dictionary, the best CDN render library, and an editor with a bookmarkable Rendered toggle when there is no fragment.
  2. Second prompt: "Safe from XSS? We could run in an iframe sandbox like we do on the SVG markdown render project" — Simon pointed it at his existing tool's <iframe sandbox> approach.
  3. Third prompt: "We should have an option to load latex from a URL (or gist) similar to the markdown SVG tool".
  4. Open the page, paste LaTeX, switch to Rendered, and share the address-bar link; the recipient sees the same content.
  5. Optional: add an llms.txt (as Tiny Little Office does) so other agents know how to construct renderable links.

Prompt

I want a tool where you can link to it with a fragment hash #code-here and it treats that code as LaTeX and renders it really effectively. It should also support a prefix which means the latex is compressed with a custom compression dictionary optimized for latex. Use the best available latex render library from a CDN (consider WebAssembly too). If loaded without a fragment hash show an editor you can paste latex into and then toggle to a "rendered" view. That rendered view should be bookmarkable. The #fragment view should let you switch to edit source too.

What the author says

Simon Willison (2026-10-15) published all three prompts and the Claude Code session link. The result, tools.simonwillison.net/latex, uses Pako 2.1.0, MathJax 4.1.3, faster-latex 0.3.1 and DOMPurify 3.4.16, with Edit/Rendered toggle, URL or Gist loading, and readable or compressed share links. He used Claude Code (Opus 5.5 Extra).

Our test

We did not re-run the three prompts in Claude Code. We only checked the published result: the page loads, and its source contains the implementation details the post describes.

What we found when we tried it:On 2026-10-17 tools.simonwillison.net/latex returned HTTP 200. Its source contains location.hash and a hashchange listener, pako 2.1.0 deflate/inflate calls, MathJax 4.1.3, DOMPurify, an iframe with sandbox="allow-scripts", and Gist-loading logic.

How to get in

English blog and tool page. Reproducing needs Claude Code, though the prompt works as a spec for any coding agent.

Who it suits

Anyone building quick no-backend share tools (formulas, diagrams, checklists, small docs), and people learning to iterate prompts in rounds.

Pitfalls

  • Rendering user-supplied content risks XSS — Simon spent a whole prompt on an iframe sandbox; don't skip it.
  • Prompts two and three reference tools already in Simon's repo; without similar references, spell out the sandbox and Gist-loading requirements.
  • Everything lives in the URL, so long documents mean long links — hence the compression prefix.
  • Fragments aren't sent to the server but do persist in browser history and wherever you share them; avoid sensitive content.

More in this category