AIKIDO-2026-570094

TL;DR

PreformattedConverter and CodeConverter entity-decode the contents of <pre> and <code> elements and wrap the decoded text in a Markdown code fence or span chosen without regard to that content, so the content can end up outside the block or span and be parsed as ordinary Markdown. Because the content is already entity decoded, text that was inert in the input HTML becomes raw HTML in the Markdown output. HTML Purifier and the strip_tags/remove_nodes options do not prevent this since the payload contains no tags when inspected. The fix sets fence length and placement from the content so decoded text cannot escape its code block or span.