MDX to Markdown

Overview

Use this skill to produce plain Markdown from MDX without requiring the target renderer to understand MDX imports, JSX components, Rspress anchors, or code-block metadata.

Workflow

1. Identify the source .mdx and target .md path. If the user gives only a source path, default the output to the same path with .md.

2. Inspect the MDX for non-Markdown constructs:

   • YAML frontmatter.
   • top-level import lines.
   • custom components such as <BlogAuthors /> and <PackageManagerTabs />.
   • heading anchors like ### Title \{#title}.
   • standalone HTML anchors like <a id="title"></a>.
   • fenced code block titles such as a JavaScript fence with title="env.js".
   • HTML image wrappers such as centered <div><img ... /></div>.

3. Run the helper script when the requested conversion matches the default rules:

   node /path/to/skills/mdx-to-markdown/scripts/convert_mdx_to_markdown.mjs input.mdx output.md
   

4. Review the result with rg and a short diff. Confirm that no MDX-only syntax remains.

5. Apply any user-requested follow-up edits, such as removing a generated table of contents or changing link policy.

Default conversion rules

• Drop YAML frontmatter and top-level MDX imports.

• Remove <BlogAuthors />.

• Expand <PackageManagerTabs command="add pkg -D" /> to a bash code block using pnpm add pkg -D.

• Convert centered HTML image blocks into Markdown image syntax.

• Remove Rspress heading anchor suffixes while keeping the heading text.

• Remove standalone <a id="..."></a> anchor lines.

• Remove table-of-contents links that target in-page anchors, preserving the list text.

• Convert code-fence titles or standalone bold filename captions into the first line of the code block as a comment, for example:

  // hello.js
  export function greet(name) {
    return `Hello, ${name}!`;
  }
  
  console.log(greet('world'));
  

• Convert code blocks that contain markers such as // [!code highlight] to diff code blocks, remove the marker, and prefix highlighted lines with +.

• Prefix Markdown links whose target starts with / using the current repository's site domain. Infer the product from the input file's Git repository, remote, package name, or path segment:

  • rspack -> rspack.rs
  • rsbuild -> rsbuild.rs
  • rspress -> rspress.rs
  • rslib -> rslib.rs
  • rsdoctor -> rsdoctor.rs
  • rstest -> rstest.rs

• Add /zh to the prefix when the MDX path contains a zh path segment. For example, a Chinese Rspack document converts /guide/optimization/tree-shaking to https://rspack.rs/zh/guide/optimization/tree-shaking.

• Use --link-prefix <prefix> only when the inferred domain is wrong or the user explicitly requests a custom prefix.

Validation checklist

Run targeted checks after conversion:

rg -n '<BlogAuthors|PackageManagerTabs|title="|\\{#|^<a id="|\\]\(/|!code highlight|^\*\*[^*]+\*\*$' output.md
rg -n '^```' output.md | wc -l

The first command should return no matches unless the user intentionally kept that syntax. The fence count should be even.

Resources

• scripts/convert_mdx_to_markdown.mjs: deterministic converter for the common Rspack/Rstack release-blog MDX pattern.