Skip to main content
Back

Nuwa-skill Hits 28K Stars Turning Public Figures Into Runnable AI Thinking Frameworks

Nuwa.skill is an open-source agent skill that distills the cognitive frameworks of any public figure into a queryable AI perspective, crossing 28k GitHub stars.

Repo, Full Access
  • What it is: Nuwa.skill is an open-source agent skill that distills any public figure's cognitive frameworks into a queryable AI perspective skill.
  • Traction: 28,000+ GitHub stars and ~4,000 forks; adopted as a default skill by Tencent, Kimi, and Zhipu AI agent products.
  • How it works: 6 parallel research agents scrape 40+ sources, apply triple-verification to extract mental models, then validate the output against known answers before shipping a SKILL.md file.
  • 15 pre-built skills included: Karpathy, Munger, Feynman, Naval, Taleb, Jobs, Musk, Paul Graham, Ilya Sutskever, and more -- all scoring A-grade on an independent fidelity benchmark.
  • Cross-platform: Built on the open Agent Skills standard; runs on Claude Code, Codex, Cursor, Gemini CLI, and 50+ runtimes via npx skills add alchaincyf/nuwa-skill.
  • Honest limits: Cannot distill intuition, captures only public information, and is frozen at research time -- each skill explicitly documents what it cannot do.

Nuwa.skill is an open-source tool that does something deceptively simple: you give it a name, and it builds a runnable AI skill that thinks the way that person thinks. Not their quotes, not their biography, but their actual decision-making frameworks, mental models, and reasoning patterns. The repo has crossed 28,000 GitHub stars and nearly 4,000 forks, and has been adopted as a default skill by agent products from Tencent, Kimi, and Zhipu AI.

The problem it's solving

Most people who want to "think like Munger" or "reason like Feynman" end up with one of two things: a list of their famous quotes, or a chatbot doing a bad impression. The key insight behind Nuwa is turning "distilling how a person thinks" into a repeatable, structured workflow. If that works, the value goes far beyond entertaining character prompts. It means taking someone's judgment framework, priorities, common heuristics, and communication habits and turning them into a skill that can be called again and again.

The predecessor project, colleague-skill, distills what a person does. Nuwa distills how a person thinks. The distinction matters. Work habits can be documented. But the cognitive framework that makes Munger and Musk reach different conclusions from the same data cannot be captured in a process doc.

Built on a cross-platform open standard

Agent Skills is an open specification, available at agentskills.io, that defines a portable format for packaging procedural knowledge for AI agents. The format originated inside Claude Code in October 2025, but Anthropic spun it out as an independent, vendor-neutral standard with a published specification and reference SDK, explicitly modeled on how the Model Context Protocol became the de facto standard for tool-calling. Within 48 hours of the open standard launch, Microsoft had wired Agent Skills into VS Code via Copilot, and OpenAI had added support to both ChatGPT and Codex CLI.

Nuwa is an Agent Skill that extracts cognitive frameworks -- mental models, decision heuristics, expression DNA -- from any public figure into a runnable perspective skill. It works in Claude Code, Codex, Cursor, OpenClaw, Hermes Agent, CodeBuddy, Workbuddy, Gemini CLI, OpenCode, and 50+ skills-compatible runtimes. Because it's built on the open standard, the same skill file works everywhere without modification.