Integrate Markdown Pipelines

This skill builds on configure-selective-highlighting. Read it first for registry and bundle behavior.

Integrate Markdown Pipelines

Setup

Connect TanStack Markdown without duplicating its <pre><code> wrapper:

import { createTanStackMarkdownHighlighter } from '@tanstack/highlight/markdown'
import { highlighter } from './highlight'

export const highlightMarkdownCode =
  createTanStackMarkdownHighlighter(highlighter)

Render a parsed fence directly when the application owns Markdown traversal:

import { renderCodeFence } from '@tanstack/highlight/markdown'
import { highlighter } from './highlight'

export function renderFence(code: string, lang: string, meta?: string) {
  return renderCodeFence(
    {
      code,
      lang,
      meta,
    },
    highlighter,
  )
}

The result includes copyText, htmlMarkup, lang, title, tokens, decorations, and lineNumbers.

Core Patterns

Theme TanStack Markdown containers

import { createThemeCss } from '@tanstack/highlight/theme'
import { githubDarkTheme } from '@tanstack/highlight/themes/github-dark'
import { githubLightTheme } from '@tanstack/highlight/themes/github-light'

export const css = createThemeCss({
  light: githubLightTheme,
  dark: githubDarkTheme,
  lightSelector: '.markdown-renderer',
  darkSelector: '.dark .markdown-renderer',
  codeBlockSelector: '.markdown-renderer pre.tm-code',
  lineNumbersSelector: '.markdown-renderer .tm-code--line-numbers',
})

The adapter returns escaped inner markup. TanStack Markdown owns the outer elements and adds tm-code--line-numbers when requested.

Transform MDAST before remark-rehype

import rehypeStringify from 'rehype-stringify'
import remarkParse from 'remark-parse'
import remarkRehype from 'remark-rehype'
import { remarkHighlightCodeBlocks } from '@tanstack/highlight/remark'
import { unified } from 'unified'
import { highlighter } from './highlight'

const processor = unified()
  .use(remarkParse)
  .use(remarkHighlightCodeBlocks, {
    highlighter,
    lineNumbers: true,
  })
  .use(remarkRehype)
  .use(rehypeStringify)

export async function renderMarkdown(markdown: string) {
  return String(await processor.process(markdown))
}

The plugin attaches structured HAST data to replacement MDAST nodes; raw HTML is not required.

Transform existing HAST code blocks

import rehypeParse from 'rehype-parse'
import rehypeStringify from 'rehype-stringify'
import { rehypeHighlightCodeBlocks } from '@tanstack/highlight/rehype'
import { unified } from 'unified'
import { highlighter } from './highlight'

const processor = unified()
  .use(rehypeParse, { fragment: true })
  .use(rehypeHighlightCodeBlocks, {
    highlighter,
    lineNumbers: false,
  })
  .use(rehypeStringify)

export async function highlightHtml(html: string) {
  return String(await processor.process(html))
}

The Rehype adapter reads <pre><code class="language-*"> and skips blocks already marked th-code.

Parse standard fence annotations

import { parseCodeFenceMeta } from '@tanstack/highlight/markdown'

export const meta = parseCodeFenceMeta(
  'title="App.tsx" {2,4-6} ins={8} error={11} lineNumbers',
)

Line annotations are one-based and inclusive; the parsed classes remain application-styled.

Common Mistakes

HIGH Confusing Markdown source with adapters

Wrong:

import { createHighlighter } from '@tanstack/highlight/core'
import { markdown } from '@tanstack/highlight/languages/markdown'

export const highlighter = createHighlighter({ languages: [markdown] })

Correct:

import { remarkHighlightCodeBlocks } from '@tanstack/highlight/remark'
import { highlighter } from './highlight'

export const highlightCode = remarkHighlightCodeBlocks({ highlighter })

The Markdown language highlights Markdown source; the adapters transform fenced blocks in a document tree.

Source: docs/guides/markdown-pipelines.md

CRITICAL Running Remark after HAST conversion

Wrong:

import remarkRehype from 'remark-rehype'
import { remarkHighlightCodeBlocks } from '@tanstack/highlight/remark'
import { unified } from 'unified'
import { highlighter } from './highlight'

export const processor = unified()
  .use(remarkRehype)
  .use(remarkHighlightCodeBlocks, { highlighter })

Correct:

import remarkRehype from 'remark-rehype'
import { remarkHighlightCodeBlocks } from '@tanstack/highlight/remark'
import { unified } from 'unified'
import { highlighter } from './highlight'

export const processor = unified()
  .use(remarkHighlightCodeBlocks, { highlighter })
  .use(remarkRehype)

The Remark adapter must see MDAST code nodes before remark-rehype converts the tree.

Source: docs/guides/markdown-pipelines.md

HIGH Enabling raw HTML without needing it

Wrong:

import {
  remarkCodeNodeToHtml,
  type RemarkCodeNode,
} from '@tanstack/highlight/remark'
import { highlighter } from './highlight'

export function transform(node: RemarkCodeNode) {
  return remarkCodeNodeToHtml(node, { highlighter })
}

Correct:

import {
  remarkCodeNodeToMdast,
  type RemarkCodeNode,
} from '@tanstack/highlight/remark'
import { highlighter } from './highlight'

export function transform(node: RemarkCodeNode) {
  return remarkCodeNodeToMdast(node, { highlighter })
}

The structured result already carries hName, hProperties, and hChildren.

Source: src/remark.ts

HIGH Passing an undiscoverable Rehype shape

Wrong:

<code data-language="tsx">const node = &lt;Button /&gt;</code>

Correct:

<pre><code class="language-tsx">const node = &lt;Button /&gt;</code></pre>

The Rehype adapter discovers a code child under pre and reads its first language-* class.

Source: src/rehype.ts

CRITICAL Mixing exact and block-normalized APIs

Wrong:

import { renderCodeFence } from '@tanstack/highlight/markdown'
import { highlighter } from './highlight'

export const serverHtml = highlighter.highlightToHtml('const value = 1\n', {
  lang: 'ts',
})
export const clientHtml = renderCodeFence(
  { code: 'const value = 1\n', lang: 'ts' },
  highlighter,
).htmlMarkup

Correct:

import { renderCodeFence } from '@tanstack/highlight/markdown'
import { highlighter } from './highlight'

export const serverHtml = renderCodeFence(
  { code: 'const value = 1\n', lang: 'ts' },
  highlighter,
).htmlMarkup
export const clientHtml = renderCodeFence(
  { code: 'const value = 1\n', lang: 'ts' },
  highlighter,
).htmlMarkup

highlight() preserves exact input, while fence and block-data helpers trim trailing block whitespace.

Source: docs/guides/ssr-and-client.md

HIGH Tension: structured output versus raw HTML

Prefer structured MDAST/HAST transforms inside Markdown pipelines. Use generated htmlMarkup only at an application boundary that deliberately inserts escaped highlighter output.

See also: integrate-framework-code-blocks/SKILL.md - direct components own their HTML insertion boundary.

HIGH Tension: exact source versus block normalization

Use the same block-level helper on server and client. Mixing core highlight() with Markdown helpers changes trailing whitespace behavior.

See also: integrate-framework-code-blocks/SKILL.md - hydrated output must use matching inputs and options.

References

• Direct fences and HAST helpers
• Remark adapter
• Rehype adapter

See also: theme-and-annotate-code/SKILL.md - fence metadata maps to titles, line numbers, and decorations.