End-to-end pipeline that takes a blog article and produces a complete 微信公众号 publishing package: WeChat-formatted content (copied to clipboard) + cover image. Optionally includes fetching from URLs, downloading embedded images, and translation for non-Chinese sources. Use this skill when the user wants multiple publishing steps handled together — formatting + cover, or translate + format + cover. Triggers on: '转公众号格式并配图', '公众号排版+封面', 'blog-to-wechat-pipeline', '翻译后发公众号', '做成公众号发布素材', '和上面一样的流程', '走一下公众号流程', '翻译成中文放进blog', '把这个链接翻译发公众号', or any bundled article-to-WeChat workflow. Also triggers when user shares a URL and mentions both translation and blog/公众号. If the user only needs one step, prefer the specific skill: wechat-md (formatting), baoyu-cover-image (cover), or baoyu-translate (translation).
Fetch source → download images → (optionally translate) → save as blog post → format for 微信公众号 → generate cover image → report.
The user wants two or more of these for a blog article:
If they only want one, use the specific skill directly:
wechat-mdbaoyu-cover-imagebaoyu-translate| Signal | Phases |
|---|---|
| User provides a URL | Fetch → rest of pipeline |
| Source is in a foreign language | Fetch → Translate → Blog post → rest |
| Source article is already a local Chinese markdown |
| Skip fetch/translate → Blog post (if needed) → rest |
| User says "翻译" / "translate" | Include translation regardless |
| User says "放进 blog" / "存到博客" | Include blog post creation |
| User says "公众号排版" / "微信格式" | Include WeChat formatting |
| User says "封面" / "配图" / "cover" | Include cover image generation |
When the user provides a URL instead of a local file, fetch the article content first.
https://r.jina.ai/<url> — returns clean markdowndefuddle skill — good for standard web pages, strips navigation/adsagent-reach skill — for platform-specific content (Twitter/X, 小红书, YouTube, etc.)webfetch — last resort, may include navigation clutter| Platform | Best method | Gotchas |
|---|---|---|
| Twitter/X long posts | Jina Reader (r.jina.ai) | Thread posts need full unrolling; xreach requires auth |
| Blog/article sites | Jina Reader or defuddle | Standard approach works well |
| WeChat articles | agent-reach (微信公众号 channel) | May need browser session |
| YouTube | agent-reach (transcript extraction) | Video → text transcription |
Save the fetched raw content to translate/<slug>.md as a working copy. This preserves the original for reference throughout the pipeline.
Source articles often contain images hosted on external CDNs. These must be downloaded before translation or blog post creation, because the blog needs local image paths.
 patterns)src/content/posts/_images/<post-title>/
.md)【译】 prefix in the folder namecurl or webfetch to the image directory
image-01-<brief-description>.jpg
Only run when translation is needed. Follow the baoyu-translate skill — check its EXTEND.md for the user's preferred translation mode (quick/normal/refined).
The baoyu-translate EXTEND.md typically specifies default_mode: refined for this user, which means a multi-pass workflow:
01-analysis.md02-prompt.md03-draft.md04-critique.md05-revision.mdSave translation artifacts to translate/<slug>-zh-CN/.
<!--more--> after the source attribution block (original URL, author, date)Save the translated (or original Chinese) article as a blog post.
The blog is an Astro project at /Users/tangcheng/Documents/Projects/blog/.
Posts directory: src/content/posts/<YYYY>/ (grouped by year)
Image directory: src/content/posts/_images/<post-title>/
.md_images/【译】Full Title Here/Content schema (from src/content.config.ts):
---