When AI Learns to Think: Deep Dive into nuwa-skill
When AI Learns to Think: Deep Dive into nuwa-skill
From "Resurrecting Colleagues" to "Cognitive OS Extraction"
"The next employee you want to distill doesn't have to be a colleague." — Huashu
Introduction: An Experiment in Cognitive Immortality
In April 2026, a GitHub project called "colleague-skill" went viral, gaining 6,800+ stars in three days. It allowed users to distill a departing colleague's work style and knowledge into an AI Skill.
But Huashu (花叔) asked a more ambitious question: "Since we can distill people, why stop at colleagues? Distill the best minds in every field."
This project is nuwa-skill.
Chapter 1: The Creator — Huashu
Huashu is an AI Native Coder, indie developer, and content creator with 300K+ followers. His代表作 (masterpiece) "小猫补光灯" (Kitten Fill Light) reached #1 on the AppStore paid charts.
Before nuwa-skill, he had already open-sourced huashu-skills — 21 content creation skills covering the entire workflow from ideation to publishing.
Chapter 2: Core Philosophy — From Roleplay to Cognitive Framework
2.1 Not Roleplay
Huashu emphasizes: "This is not roleplay. Jobs uses mental models like 'Focus means saying no'. Naval uses 'Desires are contracts'. Musk uses 'First principles'. They're not reciting quotes — they're using cognitive frameworks to analyze your problems."
2.2 Five-Layer Cognitive Extraction
| Layer | Name | Description |
|---|---|---|
| 5 | Honest Boundaries | What can't be distilled? Intuition, inspiration |
| 4 | Anti-patterns/Values | What won't they do? Red lines |
| 3 | Decision Heuristics | Rules for making choices |
| 2 | Mental Models | Core cognitive frameworks |
| 1 | Expression DNA | Tone, rhythm, word preferences |
Chapter 3: Technical Architecture
3.1 Six-Parallel Collection
- Books & Essays
- Podcasts & Interviews
- Social Media
- Critic Perspectives
- Decision Records
- Life Timeline
3.2 Triple Verification
- Cross-domain verification: Appears in 2+ domains
- Predictive power: Can infer positions on new issues
- Exclusivity: Not generic "common sense"
Chapter 4: Available Skills — 13+1 Cognitive Profiles
| Skill | Persona | Domain |
|---|---|---|
| Paul Graham | YC Founder | Startup/Writing |
| 张一鸣 | ByteDance CEO | Product/Organization |
| 张雪峰 | Education Influencer | Career Planning |
| Karpathy | AI Researcher | AI/Education |
| Jobs | Apple Co-founder | Product/Design |
| Musk | Tesla/SpaceX CEO | Engineering/First Principles |
| Munger | Berkshire Hathaway | Investment/Inversion |
| Naval | Angel Investor | Wealth/Leverage |
| Taleb | Statistician | Risk/Antifragility |
| Feynman | Physicist | Learning/Teaching |
| MrBeast | YouTuber | Content Creation |
| Trump | Business/Politics | Negotiation/Power |
| X-Mentor | Topic Skill | Twitter/X Growth |
Chapter 5: Comparison with colleague-skill
| Dimension | colleague-skill | nuwa-skill |
|---|---|---|
| Target | Your colleagues | Top minds globally |
| Data Source | Private data | Public sources |
| Content | Work habits | Cognitive frameworks |
| Depth | Application layer | OS layer |
| Privacy Risk | High | Low |
As Huashu puts it: "colleague-skill distills what people do. nuwa-skill distills how people think."
Chapter 6: Six-Dimension Quality Assessment
| Dimension | Score |
|---|---|
| Structural Integrity | 9/10 |
| Instruction Clarity | 9/10 |
| Practicality | 9/10 |
| Reproducibility | 8/10 |
| Professional Depth | 9/10 |
| Differentiation | 10/10 |
| Total | 8.8/10 (Grade A) |
Chapter 7: Usage Examples
Startup Decision (PG Perspective)
User: Should I build an AI customer service tool for SMBs? PG: You're asking "should I?" but I need more info: 1. Do YOU have pain points with SMB customer service? 2. What are existing players doing wrong? 3. Can you build a weekend prototype? Startup ideas aren't thought up — they're discovered. Problems hide in things that make you curse daily.
Career Choice (Naval Perspective)
User: I'm纠结 about an offer — 50% raise but company might not survive 2 years Naval: You've listed two desires: more money, and security. Each desire is a contract with unhappiness. Ask yourself: If this company dies in 2 years, will the skills and network you gained be worth the risk?
Chapter 8: Limitations
- Data cutoff: Each skill is a snapshot at research time
- Public ≠ Private: Can only extract from public expressions
- Intuition can't be distilled: Frameworks yes, inspiration no
Chapter 9: Future Directions
- Automated distillation tools
- Real-time update mechanisms
- Personalized fusion skills
- Enterprise cognitive asset libraries
- Multi-modal interactions
Chapter 10: Conclusion
nuwa-skill is one of the most innovative AI Skills projects of 2026. Its core value:
- Methodological innovation: From roleplay to cognitive extraction
- Systematic approach: Five layers, six sources, triple verification
- Practicality: 13 ready-to-use skills
- Open source: Fully transparent methodology
- Chinese context: Native localization
Get Started
# Install nuwa-skill npx skills add alchaincyf/nuwa-skill # Install specific persona skills npx skills add alchaincyf/paul-graham-skill npx skills add alchaincyf/zhang-yiming-skill npx skills add alchaincyf/steve-jobs-skill
References
Author: SkillScout (CMO @ skillsAgent.org)
Last Updated: April 9, 2026