WeChat Article Layout Assistant

What It Solves

WeChat Official Account editors strip or ignore parts of HTML, causing copied layouts to collapse. This skill normalizes Markdown, docx, PDF, plain text, or rich text into a stable article structure, then generates an inline-style HTML fragment suitable for pasting into the editor. It targets visual hierarchy for headings, quotes, code blocks, images, and emphasis without relying on unsupported CSS features.

How It Works

The skill is built around a theme component library plus a shared component increment library. It first reads the article and infers its type, selects or recommends a theme from the theme index, loads the theme file and common components, and assembles HTML using an “article type to component recipe” approach. During generation it handles:

• Section numbering: assigns 01, 02, 03 in ## order
• Keyword underlining: marks 1–3 key phrases in body paragraphs
• Lead highlight, digest, signature placeholders: builds intro card, selected takeaways, and author signature area
• Punctuation and code blocks: uses full-width Chinese punctuation and code blocks for code, commands, or prompts

It typically outputs a clean HTML fragment file and a preview page with a copy button. The preview page only wraps the copy action; the copied content remains a compliant section fragment, avoiding extra document shells in the editor.

Scope and Caveats

This is useful for stable WeChat long-form articles, tutorials, tool roundups, opinion pieces, and case studies; it is not intended for pages that depend on complex interactions, external fonts, grid layouts, animations, or scripts. Theme output should respect platform limits: avoid class/id, position:fixed/absolute, float, @media, @keyframes, display:grid, CSS variables, or external CSS. Critical styles should be inline, and text nodes should be wrapped with span; otherwise pasting may lose styling. Images should not default to width:100%; small images should keep original size and center. Image captions should only be generated when source data actually contains caption text. If the original already has an author signature or a “like, watch, share” block, merge it into the single closing signature area rather than duplicating CTAs.