docs(site): count imported code toward the code-block goal check

[reemsabawi-mystenlabs]

Description

A goal requirement of pattern: '```' counts backticks in the body. <ImportContent … mode="code"> renders a code block from a real source file without any backticks, so a page that imports its examples scores zero on a check it should pass, while an inline copy of the same snippet passes.

Every import-based example page fails Has code blocks for setup, source, and output: found 0, need >= 3 today.

Why this matters beyond the score

This scoring is what produced padding the repo has already had to fix. examples/javascript.mdx and examples/move.mdx each carried their intro sentence three times, which #3637 removed:

The following Move example showcases how to import and use Walrus onchain objects.

The following Move example showcases how to import and use Walrus onchain objects.

The following Move example showcases how to import and use Walrus onchain objects.

Those pages are one sentence plus an imported source file, so they could not clear min_words: 100 on their own. The checks reward pasting a copy over importing the source, which is the opposite of the house preference and the opposite of what review asks for.

Change

Each <ImportContent … mode="code"> counts as one complete fence pair for the '```' pattern, so importing a snippet never scores worse than pasting it. The check is otherwise untouched.

Test plan

node src/scripts/audit-docs.mjs from docs/site, before and after:

Page	Before	After
examples/python.mdx	found 6	found 6, unchanged
examples/walrus-relay.mdx	found 6	found 6, unchanged
examples/browser-and-mobile.mdx	found 14	found 14, unchanged
examples/javascript.mdx	found 0	found 2
examples/move.mdx	found 0	found 2
examples/checkpoint-data.mdx	found 0	found 2

Counts only increase, so no page that passed before can start failing.

The three pages that still sit below the threshold of 3 on main reach it once their open PRs land, which is where their additional imports live: examples/javascript.mdx and examples/checkpoint-data.mdx in #3627, and examples/data-marketplace.mdx in #3626.

Note on the related word-count check

min_words has the same blind spot: it measures prose, so an import-heavy page reads as short however good the imported code is. I have not changed it here, because unlike the fence count there is a real editorial argument that a page should carry enough prose to explain what its example does. The pages fixed in #3637 now clear it on real content rather than repetition. Raising it as a known adjacent issue rather than folding it into this change.

Release notes

• Storage node:
• Aggregator:
• Publisher:
• CLI:

[github-actions]

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-08-17 20:01 UTC

[github-actions]

📋 afdocs check results

URL: https://MystenLabs.github.io/walrus/pr-preview/pr-3641/

Running checks on mystenlabs.github.io/walrus/pr-preview/pr-3641/...

Agent-Friendly Docs Check: https://MystenLabs.github.io/walrus/pr-preview/pr-3641/
Timestamp: 8/13/2026, 5:13:32 PM

content-discoverability
  ✗ llms-txt-exists: No llms.txt found at any candidate location (https://MystenLabs.github.io/walrus/pr-preview/pr-3641/llms.txt, https://mystenlabs.github.io/llms.txt, https://mystenlabs.github.io/docs/llms.txt)
      Learn more: https://agentdocsspec.com/spec/#llms-txt-exists
  ○ llms-txt-valid: Skipped: dependency check did not pass
  ○ llms-txt-size: Skipped: dependency check did not pass
  ○ llms-txt-links-resolve: Skipped: dependency check did not pass
  ○ llms-txt-links-markdown: Skipped: dependency check did not pass
  ✗ llms-txt-directive-html: Could not test any pages; 1 failed to fetch
      Learn more: https://agentdocsspec.com/spec/#llms-txt-directive-html
  ✗ llms-txt-directive-md: Could not fetch markdown for any of 1 pages; 1 had no markdown version
      Learn more: https://agentdocsspec.com/spec/#llms-txt-directive-md

markdown-availability
  ✗ markdown-url-support: No pages support .md URLs (0/1 tested)
      Learn more: https://agentdocsspec.com/spec/#markdown-url-support
  ✗ content-negotiation: Server ignores Accept: text/markdown header (0/1 pages return markdown)
      Learn more: https://agentdocsspec.com/spec/#content-negotiation

page-size
  ✓ rendering-strategy: All 1 pages contain server-rendered content
  ○ page-size-markdown: Skipped: dependency check did not pass
  ✓ page-size-html: All 1 pages under 50K chars (median 21K HTML → 809 markdown (96% boilerplate))
  ⚠ content-start-position: 1 of 1 pages have content starting at 10–50% (worst 42%)
      Learn more: https://agentdocsspec.com/spec/#content-start-position

content-structure
  ✓ tabbed-content-serialization: No tabbed content detected across 1 pages
  ✓ section-header-quality: No tabbed content found; header quality check not applicable
  ○ markdown-code-fence-validity: Site does not serve markdown content; nothing to analyze

url-stability
  ✓ http-status-codes: All 1 pages return proper error codes for bad URLs
  ✓ redirect-behavior: No redirects detected across 1 pages

observability
  ○ llms-txt-coverage: Skipped: dependency check did not pass
  ○ markdown-content-parity: Skipped: dependency check did not pass
  ✓ cache-header-hygiene: All 1 endpoints have appropriate cache headers

authentication
  ✓ auth-gate-detection: All 1 pages are publicly accessible
  ○ auth-alternative-access: All docs pages are publicly accessible; no alternative access paths needed

Summary
  8 passed, 1 warnings, 5 failed, 9 skipped (23 total)

Full spec: https://agentdocsspec.com/spec/