Overview

HTML → markdown content extraction. Picks the article root from messy real-world HTML, strips noise, converts structural tags to markdown, and decodes entities. Intended output is for a markdown renderer or pager.

• Strips noise tags entirely along with their bodies (<script>, <style>, <nav>, <header>, <footer>, <aside>, <noscript>, <iframe>, <form>, <button>, <svg>).

• Converts <a href>...</a> to [text](url) so the markdown pager renders it as a clickable link.

• Converts <h1>..<h6> to #/##/... headings.

• Converts <pre>...</pre> to fenced code blocks; <code> to backticks.

• Inserts blank-line separators around paragraphs, divs, list items.

• Decodes the common HTML named entities + numeric refs.

• Caps output at 256 KB with a "… [truncated]" suffix.

The implementation is intentionally regex-based — real HTML is rarely well-formed and a strict parser would reject pages that pragmatic readers should still display. The risk surface is small: the output goes only into a markdown pager, so even pathological HTML can't escape its sandbox.

Functions