Finished writing, then spent half an hour fighting layout: mismatched titles, cramped code blocks, drifting images. You meant to publish; you became a layout worker.
We ran the open-source project gzh-design-skill (a WeChat Official Account layout helper) on our machine.
Results first
- Built a layout file that passed the tool's check
- Generated a preview page with a copy button
- Intentionally inserted three bad patterns — the tool reported three errors
Important: "Compliant" here means the layout code passed this tool's checker. It is not WeChat content moderation and not an official WeChat certification.
A preview that opens is not the same as pasting into the Official Account editor without breakage. Treat those as two separate checks.
Repo
https://github.com/isjiamu/gzh-design-skill
It takes Markdown-style drafts (hash headings, simple lists), applies a theme, and emits HTML with inline styles (font size, color, spacing on the tags themselves) so you can paste into the WeChat editor.
What we actually verified
Sample check: fully compliant. Bad sample: three errors found. Component library: no blocking failures. Combined output re-checked: still passed, then preview generated.
We did not finish the full "assistant picks theme and assembles the whole article" path this round. Do not read "sample passed" as "fully auto-layout and verified inside WeChat."
Who it's for
People who write long WeChat posts, already draft with AI assistants, and still get stuck on final layout.
Best test: your own real article (body, titles, images, code) → check → preview → paste → inspect the reader page.
— Jian AI Lab: install, run, show the measured result.
Top comments (1)
It's interesting to see how you highlighted the distinction between passing layout checks and actual content moderation within WeChat. This is a crucial point often overlooked, especially when developers assume that passing a tool's checks guarantees a flawless user experience. One improvement idea might be to integrate more comprehensive error handling that can provide contextual feedback for the user as they draft, thereby minimizing layout frustrations. If you're considering further development on enhancing the tool's capabilities or expanding its feature set, I’d be happy to explore a paid collaboration to support that effort. What additional features do you envision for the next iteration?