Markdown Rendering Regression

Markdown Rendering Regression

Purpose

Catch public documentation regressions that ordinary builds miss. Build success does not prove Markdown rendered correctly or that generated copy is safe for readers.

What To Check

• Raw Markdown syntax visible in rendered pages: **bold**, ](link), fenced code markers, frontmatter delimiters.
• Placeholder copy: _No source outputs were available_, _No structured content available_, TODO/filler markers.
• Internal process language: wave labels, current run/build labels, raw output language, local machine paths.
• Artifact wording: use Source artifact, Open artifact, or Download; do not expose Raw artifact path.
• UI icon policy: verify rendered public HTML does not expose emoji icons where the design system bans them.
• Frontmatter descriptions: sidebar cards and document rails often expose frontmatter directly, so scan full visible text, not just Markdown bodies.

Default Command

Use the helper script when available:

node ~/.agents/skills/markdown-rendering-regression/scripts/verify-public-html.mjs --dist dist

For repo-local copies:

node scripts/verify-public-html.mjs

Implementation Notes

• Run after the static build has produced dist/.
• Strip <script> and <style> before visible-text checks.
• Scan all generated HTML, not only known docs pages.
• Report exact file and banned label for every failure.
• Keep the ban list configurable by editing the script constants or passing a JSON config in future repo-specific variants.

Acceptance Examples

• A rendered page containing _No source outputs were available_ must fail.
• A rendered page containing visible **bold** must fail.
• A rendered page containing /Volumes/madara or another local absolute path must fail.
• A rendered page containing Raw artifact path must fail.