wechat-pipeline

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

WeChat Pipeline

这个 skill 用来把微信公众号内容生产流程编排成一条固定流水线。

目标是只使用现成 skill，把生产过程稳定为：

选题 -> 研究 -> 写作 -> 优化 -> 分发

这是仓库里的公开编排入口。

用户应该把它当成流程入口，而不是把内部 subskill 当成平级选择。

This skill is used to orchestrate the WeChat Official Account content production process into a fixed pipeline.

The goal is to only use off-the-shelf skills to standardize the production process as:

Topic Selection -> Research -> Writing -> Optimization -> Distribution

This is the public orchestration entry in the repository.

Users should treat it as the process entry, rather than treating internal subskills as peer options.

何时使用

When to use

当用户要在 Codex 里快速搭一套公众号内容流水线，或者明确要做下面这些事情时使用：

规划微信公众号内容生产流程
用 skill 串起选题、研究、写作、排版、发布
批量生产公众号文章
为后续扩展到其他渠道保留可替换结构

Use it when users want to quickly build a set of official account content pipelines in Codex, or explicitly need to do the following things:

Plan WeChat Official Account content production process
Use skills to connect topic selection, research, writing, typesetting, and publishing
Mass produce official account articles
Retain replaceable structure for subsequent expansion to other channels

允许使用的 Skill

Allowed Skills

主链路：

```
keyword-research
```
```
seo-content-brief
```
```
knowledge-article-calibrator
```
（可选，写前校准）
```
seo-content-writer
```
```
knowledge-article-auditor
```
（可选，写后审查）
```
baoyu-format-markdown
```
```
baoyu-markdown-to-html
```
```
baoyu-post-to-wechat
```

公众号快写链路：

```
wechat-topic-outline-planner
```
```
knowledge-article-calibrator
```
（可选，写前校准）
```
wechat-article-writer
```
```
knowledge-article-auditor
```
（可选，写后审查）
```
baoyu-format-markdown
```
```
baoyu-markdown-to-html
```
```
baoyu-post-to-wechat
```

默认优先主链路。只有在下面情况才切到快写链路：

用户明确要求“走公众号快写链路”
当前主题更强调公众号表达，而不是研究深度
主链路中的研究或写作 skill 明显不适配当前目标

Main pipeline:

```
keyword-research
```
```
seo-content-brief
```
```
knowledge-article-calibrator
```
(optional, pre-writing calibration)
```
seo-content-writer
```
```
knowledge-article-auditor
```
(optional, post-writing review)
```
baoyu-format-markdown
```
```
baoyu-markdown-to-html
```
```
baoyu-post-to-wechat
```

Official account quick writing pipeline:

```
wechat-topic-outline-planner
```
```
knowledge-article-calibrator
```
(optional, pre-writing calibration)
```
wechat-article-writer
```
```
knowledge-article-auditor
```
(optional, post-writing review)
```
baoyu-format-markdown
```
```
baoyu-markdown-to-html
```
```
baoyu-post-to-wechat
```

The main pipeline is prioritized by default. Switch to the quick writing pipeline only in the following cases:

User explicitly requests to "use the official account quick writing pipeline"
The current topic emphasizes WeChat Official Account expression more than research depth
The research or writing skills in the main pipeline are obviously not suitable for the current goal

本仓库内推荐增强 Skill

Recommended Enhanced Skills in This Repository

这条流水线允许接入本仓库的公开 skill 作为增强节点：

```
ai-workflow-brand
```
```
reader-hook
```

knowledge-article-system

（含

knowledge-article-calibrator

和

knowledge-article-auditor

）

使用原则：

如果主题方向、定位、栏目归属还不清楚，先用
```
ai-workflow-brand
```
如果文章主线没问题，但标题和开头不够有进入感，用
```
reader-hook
```
如果准备发布前需要检查是否跑偏，也回到
```
ai-workflow-brand
```
如果文章是知识类判断型内容（AI、工作流、认知、内容生产、判断力），在写作前用
```
knowledge-article-calibrator
```
做预写校准，在初稿完成后用
```
knowledge-article-auditor
```
做审查评分
如果不确定用哪个，通过
```
knowledge-article-system
```
母 skill 路由

不要要求用户直接记忆

ai-workflow-brand

的内部 subskill。对用户来说，公开入口始终是母 skill。

This pipeline allows access to public skills in this repository as enhancement nodes:

```
ai-workflow-brand
```
```
reader-hook
```

knowledge-article-system

(including

knowledge-article-calibrator

and

knowledge-article-auditor

)

Usage principles:

If the topic direction, positioning, and column attribution are still unclear, use
```
ai-workflow-brand
```
first
If the main line of the article is fine, but the title and opening are not attractive enough, use
```
reader-hook
```
If you need to check whether the content deviates from the theme before publishing, also use
```
ai-workflow-brand
```
If the article is knowledge-based judgment content (AI, workflow, cognition, content production, judgment), use
```
knowledge-article-calibrator
```
for pre-writing calibration before writing, and use
```
knowledge-article-auditor
```
for review and scoring after the first draft is completed
If you are not sure which one to use, route through the parent skill
```
knowledge-article-system
```

Do not require users to directly memorize the internal subskills of

ai-workflow-brand

. For users, the public entry is always the parent skill.

工作目录

Working Directory

工作根目录使用

<workspace_root>

。

默认值：

/root/workspace/data/wechat-pipeline

目录约定：

```
<workspace_root>/00_topics/
```
```
<workspace_root>/01_research/
```
```
<workspace_root>/02_drafts/
```
```
<workspace_root>/03_polish/
```
```
<workspace_root>/04_publish/
```
```
<workspace_root>/_state/
```

slug

默认不是必填。

如果用户没有提供

slug

，自动按下面规则生成：

YYYY-MM-DD-主题

生成规则：

日期使用当前本地日期
主题使用用户给出的主题文本
把空格替换为
```
-
```
删除明显不适合文件名的标点
保留中文或 ASCII 字母数字都可以，不需要强制转拼音

示例：

主题：
```
AI 工作流
```
自动生成：
```
2026-03-17-AI-工作流
```

固定文件契约：

选题输出：
```
<workspace_root>/00_topics/<slug>.md
```
研究输出：
```
<workspace_root>/01_research/<slug>.md
```
初稿输出：
```
<workspace_root>/02_drafts/<slug>.md
```
优化输出：
```
<workspace_root>/03_polish/<slug>.md
```
发布输出：
```
<workspace_root>/04_publish/<slug>.html
```
状态文件：
```
<workspace_root>/_state/<slug>.json
```

The working root directory uses

<workspace_root>

Default value:

/root/workspace/data/wechat-pipeline

Directory convention:

```
<workspace_root>/00_topics/
```
```
<workspace_root>/01_research/
```
```
<workspace_root>/02_drafts/
```
```
<workspace_root>/03_polish/
```
```
<workspace_root>/04_publish/
```
```
<workspace_root>/_state/
```

slug

is not required by default.

If the user does not provide

slug

, it is automatically generated according to the following rules:

YYYY-MM-DD-topic

Generation rules:

Use the current local date for the date
Use the topic text given by the user for the topic
Replace spaces with
```
-
```
Remove punctuation that is obviously not suitable for file names
Both Chinese or ASCII alphanumeric characters can be retained, no need to force conversion to Pinyin

Example:

Topic:
```
AI 工作流
```
Automatically generated:
```
2026-03-17-AI-工作流
```

Fixed file contract:

Topic selection output:
```
<workspace_root>/00_topics/<slug>.md
```
Research output:
```
<workspace_root>/01_research/<slug>.md
```
First draft output:
```
<workspace_root>/02_drafts/<slug>.md
```
Optimization output:
```
<workspace_root>/03_polish/<slug>.md
```
Publishing output:
```
<workspace_root>/04_publish/<slug>.html
```
Status file:
```
<workspace_root>/_state/<slug>.json
```

核心规则

Core Rules

每次只推进一个阶段，不要跳步。
每个阶段都要明确输入文件和输出文件。
如果上一步产物不存在，不要继续下一步。
优先复用已有产物，除非用户明确要求重写。
每完成一步，都简短汇报当前产物路径和下一步。
除非用户明确说“发布”，否则做到 HTML 产物后暂停。
如果某个 skill 不适合当前主题，要先说明原因，再切换到备选链路。
一篇文章对应一个
```
slug
```
，不要把多个主题混进同一条产线。
如果用户没传
```
slug
```
，先生成默认
```
slug
```
，再开始第一阶段。

如果用户没传

workspace_root

，默认使用

/root/workspace/data/wechat-pipeline

。

Only advance one stage at a time, do not skip steps.
Each stage must have clear input files and output files.
If the product of the previous step does not exist, do not proceed to the next step.
Prioritize reusing existing products unless the user explicitly requests a rewrite.
After completing each step, briefly report the current product path and the next step.
Unless the user explicitly says "publish", pause after generating the HTML product.
If a skill is not suitable for the current topic, explain the reason first, then switch to the alternative pipeline.
One article corresponds to one
```
slug
```
, do not mix multiple topics into the same pipeline.
If the user does not pass
```
slug
```
, generate the default
```
slug
```
first, then start the first stage.

If the user does not pass

workspace_root

, use

/root/workspace/data/wechat-pipeline

by default.

执行流程

Execution Process

阶段 0：可选定位收束

Stage 0: Optional Positioning Convergence

如果用户给的是宽泛主题，或本次内容更依赖品牌判断而不是关键词研究，先使用

ai-workflow-brand

收束定位、主线和选题方向。

这个阶段是可选增强，不替代后续正式阶段。

If the user gives a broad topic, or this content relies more on brand judgment than keyword research, first use

ai-workflow-brand

to converge the positioning, main line and topic direction.

This stage is an optional enhancement and does not replace the subsequent formal stages.

阶段 1：选题

Stage 1: Topic Selection

使用

keyword-research

。

输入：

用户给出的主题方向

如果用户没给

slug

，先根据主题自动生成。

输出：

```
<workspace_root>/00_topics/<slug>.md
```

最低要求：

给出 10 个可写选题
每个选题包含关键词、搜索意图、目标读者
给出优先级
最终收敛成 1 个最值得写的题目

Use

keyword-research

Input:

Topic direction given by the user

If the user does not provide

slug

, automatically generate it based on the topic first.

Output:

```
<workspace_root>/00_topics/<slug>.md
```

Minimum requirements:

Provide 10 writable topics
Each topic includes keywords, search intent, target readers
Provide priority
Finally converge to 1 most worth writing topic

阶段 2：研究

Stage 2: Research

默认使用

seo-content-brief

。

输入：

```
<workspace_root>/00_topics/<slug>.md
```

输出：

```
<workspace_root>/01_research/<slug>.md
```

最低要求：

文章目标
读者画像
核心论点
结构提纲
证据点或案例建议
标题方向

如果使用公众号快写链路，则在这个阶段改用

wechat-topic-outline-planner

直接生成公众号大纲。

Use

seo-content-brief

by default.

Input:

```
<workspace_root>/00_topics/<slug>.md
```

Output:

```
<workspace_root>/01_research/<slug>.md
```

Minimum requirements:

Article objectives
Reader portrait
Core arguments
Structural outline
Evidence points or case suggestions
Title direction

If using the official account quick writing pipeline, switch to

wechat-topic-outline-planner

at this stage to directly generate the official account outline.

阶段 2.5：可选写前校准

Stage 2.5: Optional Pre-writing Calibration

如果当前文章属于知识类判断型内容（AI、工作流、认知、内容生产、判断力），在进入写作前使用

knowledge-article-calibrator

做预写校准。

输入：

```
<workspace_root>/01_research/<slug>.md
```

输出：

校准结果写入研究文件末尾，或作为单独备注记录

校准内容：

误解识别（many people think X, but actually Y）
真实场景（1-3 个）
核心判断句
问题重算
建议结构
写作风险

如果校准结果显示主题太泛、没有真实场景或缺少核心判断，不要进入写作，先回到研究阶段修正。

这个阶段是可选增强。只有文章主题明确属于知识判断型内容时才触发。

If the current article is knowledge-based judgment content (AI, workflow, cognition, content production, judgment), use

knowledge-article-calibrator

for pre-writing calibration before entering writing.

Input:

```
<workspace_root>/01_research/<slug>.md
```

Output:

The calibration result is written at the end of the research file, or recorded as a separate note

Calibration content:

Misconception identification (many people think X, but actually Y)
Real scenarios (1-3)
Core judgment sentences
Problem recalculation
Suggested structure
Writing risks

If the calibration result shows that the topic is too broad, has no real scenario or lacks core judgment, do not enter writing, go back to the research stage to revise first.

This stage is an optional enhancement. It is triggered only when the article topic clearly belongs to knowledge-based judgment content.

阶段 3：写作

Stage 3: Writing

默认使用

seo-content-writer

。

输入：

```
<workspace_root>/01_research/<slug>.md
```

输出：

```
<workspace_root>/02_drafts/<slug>.md
```

最低要求：

标题有点击动机但不过度夸张
结构清晰
有案例或场景
结尾有行动建议

如果使用公众号快写链路，则在这个阶段改用

wechat-article-writer

。

Use

seo-content-writer

by default.

Input:

```
<workspace_root>/01_research/<slug>.md
```

Output:

```
<workspace_root>/02_drafts/<slug>.md
```

Minimum requirements:

The title has click motivation but is not excessively exaggerated
Clear structure
Have cases or scenarios
There are action suggestions at the end

If using the official account quick writing pipeline, switch to

wechat-article-writer

at this stage.

阶段 3.5：可选写后审查

Stage 3.5: Optional Post-writing Review

如果当前文章属于知识类判断型内容，在进入优化前使用

knowledge-article-auditor

对初稿做 20 分制审查。

输入：

```
<workspace_root>/02_drafts/<slug>.md
```

输出：

审查报告（总分、最强部分、最大问题、各项评分、必须修复、重写顺序）
如果总分低于 12，优先修复初稿，不进入格式优化

评分标准：

16-20：可以进入优化
12-15：进入优化但标注待修复项
0-11：返回写作阶段修正

这个阶段是可选增强。只有文章主题明确属于知识判断型内容时才触发。

If the current article is knowledge-based judgment content, use

knowledge-article-auditor

to review the first draft on a 20-point scale before entering optimization.

Input:

```
<workspace_root>/02_drafts/<slug>.md
```

Output:

Review report (total score, strongest part, biggest problem, score for each item, must be fixed, rewrite order)
If the total score is lower than 12, prioritize fixing the first draft, do not enter format optimization

Scoring criteria:

16-20: Can enter optimization
12-15: Enter optimization but mark items to be fixed
0-11: Return to the writing stage for revision

This stage is an optional enhancement. It is triggered only when the article topic clearly belongs to knowledge-based judgment content.

阶段 4：优化

Stage 4: Optimization

使用

baoyu-format-markdown

。

输入：

```
<workspace_root>/02_drafts/<slug>.md
```

输出：

```
<workspace_root>/03_polish/<slug>.md
```

最低要求：

补齐标题
补齐摘要
调整小标题层级
整理列表和重点强调
补齐 CTA

Use

baoyu-format-markdown

Input:

```
<workspace_root>/02_drafts/<slug>.md
```

Output:

```
<workspace_root>/03_polish/<slug>.md
```

Minimum requirements:

Complete the title
Complete the abstract
Adjust the subtitle hierarchy
Organize lists and key emphasis
Complete CTA

阶段 4.5：可选入口优化

Stage 4.5: Optional Entry Optimization

如果草稿或优化稿已经成立，但标题和开头仍然偏抽象、偏结论、缺少点击动机，则使用

reader-hook

做入口优化。

输入：

<workspace_root>/02_drafts/<slug>.md

或

<workspace_root>/03_polish/<slug>.md

输出：

更新后的标题和开头方案，写回当前稿件或作为单独优化建议记录

这个阶段不改变文章主线，只改善读者进入感。

If the draft or optimized draft is already valid, but the title and opening are still too abstract, too conclusion-oriented, and lack click motivation, use

reader-hook

for entry optimization.

Input:

<workspace_root>/02_drafts/<slug>.md

<workspace_root>/03_polish/<slug>.md

Output:

Updated title and opening plan, written back to the current manuscript or recorded as a separate optimization suggestion

This stage does not change the main line of the article, only improves the reader's entry experience.

阶段 5：分发准备

Stage 5: Distribution Preparation

使用

baoyu-markdown-to-html

。

输入：

```
<workspace_root>/03_polish/<slug>.md
```

如果用户希望在转 HTML 前做一次品牌主线检查，先使用

ai-workflow-brand

对当前稿件进行审查，再继续分发准备。

输出：

```
<workspace_root>/04_publish/<slug>.html
```

Use

baoyu-markdown-to-html

Input:

```
<workspace_root>/03_polish/<slug>.md
```

If the user wants to do a brand main line check before converting to HTML, first use

ai-workflow-brand

to review the current manuscript, then continue with distribution preparation.

Output:

```
<workspace_root>/04_publish/<slug>.html
```

阶段 6：发布

Stage 6: Publishing

只有用户明确说“发布”时，才使用

baoyu-post-to-wechat

。

输入：

```
<workspace_root>/04_publish/<slug>.html
```

输出：

已发布到微信公众号，或给出发布过程中的登录/API 选择

Only when the user explicitly says "publish", use

baoyu-post-to-wechat

Input:

```
<workspace_root>/04_publish/<slug>.html
```

Output:

Published to WeChat Official Account, or provide login/API selection during the publishing process

阶段切换条件

Stage Switching Conditions

只有满足下面条件，才能进入下一阶段：

```
00_topics/<slug>.md
```
存在，才能进入研究
```
01_research/<slug>.md
```
存在，才能进入写作
```
02_drafts/<slug>.md
```
存在，才能进入优化
```
03_polish/<slug>.md
```
存在，才能转 HTML
```
04_publish/<slug>.html
```
存在，且用户明确要求发布，才能进入公众号发布

可选增强阶段不会改变主阶段的必需产物要求。

Only when the following conditions are met can you enter the next stage:

```
00_topics/<slug>.md
```
exists before entering research
```
01_research/<slug>.md
```
exists before entering writing
```
02_drafts/<slug>.md
```
exists before entering optimization
```
03_polish/<slug>.md
```
exists before converting to HTML
```
04_publish/<slug>.html
```
exists and the user explicitly requests publishing before entering official account publishing

Optional enhancement stages do not change the required product requirements of the main stages.

批量执行建议

Batch Execution Suggestions

如果用户要做周更或批量生产，按这个节奏分批执行：

周一统一跑选题
周二统一跑研究
周三统一跑写作
周四统一跑优化
周五统一跑 HTML 和发布

If users want to update weekly or produce in batches, execute in batches according to this rhythm:

Run topic selection uniformly on Monday
Run research uniformly on Tuesday
Run writing uniformly on Wednesday
Run optimization uniformly on Thursday
Run HTML conversion and publishing uniformly on Friday