Example URL(s)

On pages like /workers/wrangler/commands/, “Copy Page” and "View Page as Markdown" return "Not Found" – check https://developers.cloudflare.com/workers/wrangler/commands/index.md

The breakage comes from the Expressive Code integration. It renders the copy-to-clipboard button by placing the raw snippet in data-code, so a command such as wrangler deploy [<SCRIPT>] [OPTIONS] ends up as:

<button … data-code="wrangler deploy [<SCRIPT>] [OPTIONS]">

node-html-parser then sees <SCRIPT and assumes a <script> tag was opened but never closed, which corrupts .sl-markdown-content and causes the 404.

Actual Behavior

Switching the parser here to parse(html, { parseNoneClosedTags: true }) makes the Markdown export succeed, but it’s a band-aid. That option instructs the parser to auto-close any tag it thinks is left hanging, so <SCRIPT> starters stay “open” until the end of the document. Because no real </SCRIPT> exists, the synthetic closing tag lands after the footer. As soon as .sl-markdown-content sits inside that synthetic wrapper, its closing </div> stops matching and the Markdown export starts including all of the footer/sidebar as well.

Expected Behavior

Perhaps a better fix would be to escape those raw tags inside data-code (and similar attributes) before they reach node-html-parser. The browser would still decode them for copy-to-clipboard, but the HTML would remain valid and the Markdown generation would stop breaking.