Inject an AI assistant with a set of writing skills in the Chinese style, following the tone of Liu Run or Luo Zhenyu, automatically generating long-form business or inspirational articles with personal flair.

Learn Liu Run’s business insights and methodology; luozhenyu-bookwriter learns Luo Zhenyu’s cognitive inspiration and business strategy. One is more about “thoroughly explaining things”, the other is more about “waking people up”. Liu Run’s approach compresses his 28 years of writing experience into 8 core principles, 3 structures, SCQA’s three types of logical momentum, the 5 Business School SCA++ template, and 7 special content blocks, and runs a 12-item quality self-check before writing.

In actual use, you just give the Agent one command: “Write a 32-point commentary about X in Liu Run’s style.” It will pick the structure, select the momentum, find material, start writing, and wait for your approval after each section. Finally, a Python script connects pandoc and XeLaTeX to output a PDF.

liangdabiao/liurun-bookwriter-skills

Source: https://github.com/liangdabiao/liurun-bookwriter-skills

liurun-bookwriter · Dual-Style Long-Form Writing Skill Set

Two skills, two styles, one goal — help you write long-form articles with a “writer’s taste.”

This project includes two complementary Chinese long-form writing skills:

Quick Selection

Not sure? Consider the core question — is your article about “solving problems” or “stimulating thinking”? For the former, choose Liu Run; for the latter, choose Luo Zhenyu.

For detailed comparison, see §III Comparison and Complementarity.

I. liurun-bookwriter · Liu Run Style Business Long-Form Text

Empathy is the most important foundational skill. In your heart, it’s not “me”, it’s “you”. — Liu Run

A complete replication of Liu Run’s (former Microsoft Strategic Cooperation Director, founder of Runmi Consulting, host of “5-Minute Business School”) 28 years of business writing experience. Distilled from his “28 Years of Writing Principles” into 8 core principles, 3 structures, SCQA logical momentum, the 5 Business School SCA++ template, and 7 callout types, designed specifically for business long-form creation.

What This Skill Can Do

Get Started in 5 Minutes

Step 1: Trigger the Skill

Directly say to the Agent:

"Write a 32-point commentary about HarmonyOS in Liu Run's style"

The Agent will automatically execute the following workflow:

``

1. Identify structure type → 32 points
2. Select SCQA variant → CSA (default)
3. Confirm topic angle → HarmonyOS 32 points
4. Research material (automatic) → Cases + data + quotes
5. Write → Follow the 8 core principles
6. QC self-check → 12 quality checks
7. Wait for your confirmation for each point ``

Step 2: Confirm After Each Section

After finishing each section/point/item, the Agent will pause and wait for your confirmation. No manual action needed.

Step 3: Export to PDF

bash python scripts/export_pdf.py article.md

8 Core Writing Principles (Core Mental Models)

For detailed explanation, see references/style-dna.md

3 Structure Blueprints

How to choose?

• Not sure → Default to 3-section (basic editor requirement)
• Many cases, includes interviews → 10-point
• Express a stance, be restrained → 32-point
• Produce 5-Minute Business School audio → 5 Business School SCA++

For detailed blueprints, see references/book-blueprints.md

3 SCQA Logical Momentum Variants

Explicit annotation method:

``markdown A Answer: Change the commission system to a bonus system. [Body content expands]

C Complication: But doing this will provoke backlash from the sales team. [Body content expands] ``

For detailed logical momentum explanation, see references/style-dna.md “Principle 4”

7 Callout Special Content Blocks

For detailed formats, see references/callout-patterns.md

Installation

Method 1: Copy to Agent skills directory

``bash

Claude Code

cp -r liurun-bookwriter ~/.claude/skills/

WorkBuddy

cp -r liurun-bookwriter ~/.workbuddy/skills/ ``

Method 2: Use as local skill directory

Directly tell the Agent the path to the skill; no copying required.

Dependencies

Platform-specific Installation

macOS

bash brew install pandoc brew install --cask mactex brew install --cask font-source-han-serif brew install --cask font-lxgw-wenkai

Linux (Ubuntu/Debian)

bash sudo apt install pandoc texlive-xetex texlive-lang-chinese fonts-noto-cjk

Windows

powershell winget install --id JohnMacFarlane.Pandoc winget install --id MiKTeX.MiKTeX

Verify Dependencies

bash python scripts/export_pdf.py --list-fonts echo "# Test" > test.md python scripts/export_pdf.py test.md

Command Line Tools

1. PDF Export

``bash

Basic usage

python scripts/export_pdf.py article.md

Specify output file

python scripts/export_pdf.py article.md -o output.pdf

Specify font

python scripts/export_pdf.py article.md –font “Source Han Serif SC”

Verbose output

python scripts/export_pdf.py article.md -v

List available Chinese fonts

python scripts/export_pdf.py –list-fonts ``

PDF Styles (optimized for business long-form text):

• Font: Source Han Serif (Windows/Linux) / LXGW WenKai (preferred) / Microsoft YaHei (fallback)
• Font size: 11pt
• Line spacing: 1.5
• Margins: 3cm

2. Structure Validation

``bash

Validate long-form structure (auto-detect 3-section/10-point/32-point/5 Business School)

python scripts/validate_structure.py article.md

Silent mode (only return exit code)

python scripts/validate_structure.py article.md –quiet ``

Automatically checks 12 items:

✅ Article type detection ✅ SCQA variant detection ✅ Paragraph count correctness ✅ Paragraph number continuity ✅ Zero tolerance for forbidden words ✅ Empathy anchor ✅ Frequency of "I" / "you" ✅ "everyone" = 0 ✅ Callout count ✅ SCQA node annotation ✅ "Final words" at paragraph end ✅ Fuzzy word check

3. Multi-Agent Collaboration

``bash

Initialize a 32-point long-form article

python scripts/agent_scheduler.py –init “32-tiao-hongmeng” –type “32条” –scqa “CSA”

Initialize a 10-point long-form article

python scripts/agent_scheduler.py –init “hua-wei-10” –type “10点式” –scqa “QSCA”

Check current status

python scripts/agent_scheduler.py –status

Get next batch of tasks

python scripts/agent_scheduler.py –next

Advance to next phase

python scripts/agent_scheduler.py –advance ``

5 Agent roles: Editor-in-Chief / Researcher / Writer / Editor / Typesetter For detailed workflow, see references/agent-protocol.md

File Structure

liurun-bookwriter/ ├── SKILL.md # Main skill definition (Agent auto-loads) ├── README.md # This file ├── LICENSE # MIT License │ ├── references/ # Reference documents (Agent consults during writing) │ ├── book-blueprints.md # 3 structure blueprints + SCQA pairings │ ├── chapter-templates.md # 3 structure templates + 5 Business School SCA++ template │ ├── style-dna.md # 8 core principles complete DNA │ ├── opening-techniques.md # 6 opening techniques │ ├── callout-patterns.md # 7 callout formats │ ├── quality-checkpoints.md # 12+10 item QC checklist │ ├── agent-protocol.md # Agent collaboration workflow │ └── research-sources.md # Business research sources │ ├── assets/ # Resource templates │ ├── templates/ │ │ ├── cover-template.md # 4 long-form cover templates │ │ └── toc-template.md # 4 structure table of contents templates │ └── schemas/ │ └── outline-schema.json # Outline data structure │ ├── scripts/ # Utility scripts │ ├── export_pdf.py # PDF export (Source Han Serif / 11pt / line spacing 1.5) │ ├── validate_structure.py # Structure validation (12 items automatic QC check) │ └── agent_scheduler.py # Multi-Agent collaboration scheduler │ └── tests/ # Tests ├── baseline-checks.md # 5 baseline failure modes ├── pressure-scenarios.md # 7 stress test scenarios └── fixtures/ └── sample-3duan.md # 3-section sample long-form article

Quality Assurance

Each long-form article automatically runs 12 QC checks:

Structure Checks (4 items)

• Article type consistent with metadata
• Correct paragraph count (3-section=5, 10-point=12, 32-point=32, 5 Business School=5)
• Paragraph ends with “Final Words” / “Conclusion” (3-section, 10-point)
• Continuous numbering without breaks

Style Checks (5 items)

• Zero forbidden words (key: “everyone” = 0)
• Frequency of “I” ≥ 3 times per thousand words
• Frequency of “you” ≥ 8 times per thousand words
• Consecutive fuzzy words ≤ 3 times
• Sentence length ≤ 45 characters

Content Checks (3 items)

• Empathy anchor within first 3 paragraphs
• Callout count ≥ 1 (Quote/Story/Example/Analogy)
• At least 1 explicit SCQA node annotation

For complete checklist, see references/quality-checkpoints.md

Scenarios Not Suitable

• Academic paper format
• Pure fiction/novel creation
• Ultra-short copy (social media, Weibo)
• Financial reports requiring complex charts
• Legal clauses, technical manuals

Example: 3-Section Long-Form Article

Refer to the complete example in tests/fixtures/sample-3duan.md. The structure is as follows:

``markdown

[Title]

Creator: Liu Run Source Column: WeChat Official Account “Liu Run” Publication Date: YYYY-MM-DD Structure Type: 3-Section SCQA Variant: ASC Target Readers: Entrepreneurs · Business Readers Core Insight: [One sentence]

§01 Introduction: Why This Article

[Empathy anchor] [Why readers should continue reading]

§02 [Body 1: First core argument]

[Start from the answer (ASC’s A)] [Support with story/example/analogy] [Conclude with a quote]

§03 [Body 2: Second core argument]

[Follow from §02’s ending] [Deeper insight] [Conclude with a quote]

§04 [Body 3: Third core argument]

[Go even deeper] [Actionable method or suggestion] [Conclude with a quote]

§05 Final Words

[Review the three arguments] [Elevate to a higher level] [Conclude with a quote / method / question] ``

Common Questions

Q: Can I write other styles of business long-form articles?

Not recommended. This skill specializes in replicating Liu Run’s style. If you want to write in other styles, turn off this skill.

Q: What if I can’t fill all 32 points?

Don’t force it. It’s better to say less. You can switch to the 3-section (5 paragraphs) or 10-point (10 points) style.

Q: How to fix garbled Chinese text in PDF?

``bash

Confirm Chinese fonts are installed

fc-list :lang=zh

macOS: brew install –cask font-source-han-serif

Linux: sudo apt install fonts-noto-cjk

``

Q: Can I quote other writers in the long-form article?

Yes, but you must indicate the source and avoid implying it’s Liu Run’s own opinion. See the “Quote Material Library” in references/research-sources.md.

Q: How should business data be handled?

Must annotate timeliness (“as of September 2024”) and source (company announcement / authoritative media / well-known research institution). See references/research-sources.md.

II. luozhenyu-bookwriter · Luo Zhenyu Style Inspirational Long-Form Text

People who do things are not armchair generals pointing fingers, but roll up their sleeves and make themselves the key variable in solving problems. — Luo Zhenyu

A complete replication of Luo Zhenyu’s (founder of Dedao App, host of “Luoji Thinking”, host of the “Friends of Time” New Year’s Eve speech, host of “Journey of Civilization”) 14 years of content creation experience. Distilled from the video scripts of “Luoji Thinking” (2012-2026), 3652 audio entries of “Luo Pang’s 60 Seconds”, three published books (“Those Who Achieve Greatness Don’t Get Entangled”, “The Method of Reading”, “Inspiration”), 11 “Friends of Time” New Year’s Eve speeches, and the “Journey of Civilization” program into 12 core principles, 3 structures, 3 SCQA variants, 7 core callouts (+6 extensions), designed specifically for cognitive/inspiration/business long-form creation. Core principle: Learn the style, don’t copy sentences; every quote/story/analogy must be original to the piece.

For detailed documentation, see luozhenyu-bookwriter/README.md

What This Skill Can Do

12 Core Writing Principles (Core Mental Models)

⚠️ Most Important: Quotes/stories/analogies must be original to the current topic, do not copy original sentences from representative works.

For detailed explanation, see style-dna.md

3 Structure Blueprints

How to choose?

• Not sure → Default to Luoji Thinking long article (TQA Trigger-Question-Answer)
• Short content, need speed → Luo Pang 60 Seconds (PEI Phenomenon-Explanation-Inspiration)
• New Year’s Eve speech, need impact → New Year’s Eve Speech (CAA Concrete-Abstract-Action)

For detailed blueprints, see book-blueprints.md

3 SCQA Logical Momentum Variants (Luo Zhenyu Specific)

Explicit annotation method:

``markdown T Trigger: When Zeng Guofan first arrived in Changsha to train troops, his Hunan soldiers were unmanageable. Instead of panicking, he first built solid camps and dug deep trenches.

[Body content expands]

Q Question: Why does “doing the task at hand” break entanglements better than “thinking about the past”?

[Body content expands]

A Answer: Because the essence of entanglement is “thinking about the past.” The antidote is to move your attention from the past back to the present.

[Body content expands] ``

7 Core Callout Special Content Blocks

There are also 6 extended callouts (SCQA nodes / SCQA variants / decision frameworks / confused wise people / binary deconstruction / counter-intuition), see callout-patterns.md.

Installation

``bash

Already in project’s .workbuddy/skills/ directory (default location)

ls d:/liurun-bookwriter-main/.workbuddy/skills/luozhenyu-bookwriter/

Or copy to another location

cp -r .workbuddy/skills/luozhenyu-bookwriter ~/.claude/skills/ ``

Directly tell the Agent the path to the skill to use it.

Dependencies

Command Line Tools

1. Structure Validation

``bash

Validate long-form structure (auto-detect Luo Pang 60 Seconds / Luoji Thinking long article / New Year’s Eve speech)

python scripts/validate_structure.py article.md

Silent mode (only return exit code)

python scripts/validate_structure.py article.md –quiet ``

Automatically checks 12 items:

✅ Article type detection ✅ SCQA variant detection ✅ Paragraph count correctness ✅ Paragraph number continuity ✅ Zero tolerance for forbidden words ✅ "Roll up sleeves" anchor ✅ Frequency of "I" / "you" ✅ "everyone" = 0 ✅ Callout count ✅ SCQA node annotation ✅ "Conclusion" at paragraph end ✅ Fuzzy word check

2. Multi-Agent Collaboration

``bash

Initialize a New Year’s Eve speech

python scripts/agent_scheduler.py –init “ai-1000days” –type “跨年演讲稿” –scqa “CAA”

Initialize a Luo Pang 60 seconds

python scripts/agent_scheduler.py –init “60s-decision” –type “罗胖60秒” –scqa “PEI”

Check current status / advance

python scripts/agent_scheduler.py –status python scripts/agent_scheduler.py –next python scripts/agent_scheduler.py –advance ``

5 Agent roles: Editor-in-Chief / Researcher / Writer / Editor / Typesetter For detailed workflow, see agent-protocol.md

3. PDF Export

``bash

Convert Markdown long-form to PDF

python scripts/export_pdf.py article.md -o article.pdf

Specify font

python scripts/export_pdf.py article.md –font “Microsoft YaHei” ``

File Structure

.workbuddy/skills/luozhenyu-bookwriter/ ├── SKILL.md # Main skill definition (Agent auto-loads) ├── README.md # This file ├── LICENSE # MIT License │ ├── references/ # Reference documents (Agent consults during writing) │ ├── book-blueprints.md # 3 structure blueprints + 3 SCQA pairings │ ├── chapter-templates.md # 3 structure templates + New Year's Eve speech SCA template │ ├── style-dna.md # 12 core principles complete DNA + generation formulas │ ├── opening-techniques.md # 6 opening techniques │ ├── callout-patterns.md # 7 core callout + 6 extended callout formats │ ├── quality-checkpoints.md # 12+10 item QC checklist │ ├── agent-protocol.md # Agent collaboration workflow │ ├── research-sources.md # Luo Zhenyu content research sources │ └── real-works-extraction.md# Feature extraction from 20 representative works │ ├── assets/ # Resource templates │ ├── templates/ │ └── schemas/ │ ├── scripts/ # Utility scripts │ ├── validate_structure.py # Structure validation (12 items automatic QC check) │ ├── agent_scheduler.py # Multi-Agent collaboration scheduler │ └── export_pdf.py # PDF export │ └── tests/ # Tests ├── baseline-checks.md ├── pressure-scenarios.md └── fixtures/ └── sample-3duan.md # Luoji Thinking long article sample

Quality Assurance

Each long-form article automatically runs 12 QC checks:

Structure Checks (4 items)

• Article type consistent with metadata
• Correct paragraph count (Luo Pang 60 Seconds = 1 paragraph / Luoji Thinking long article = 5 paragraphs / New Year’s Eve speech = multiple sub-topics)
• Paragraph ends with “Conclusion” (Luoji Thinking long article, New Year’s Eve speech)
• Continuous numbering without breaks

Style Checks (5 items)

• Zero forbidden words (key: “everyone” = 0)
• Frequency of “I” ≥ 4 times per thousand words
• Frequency of “you” ≥ 6 times per thousand words
• Consecutive fuzzy words ≤ 3 times