alchaincyf/nuwa-skill

What it solves

Nuwa-skill allows users to "distill" the thinking patterns and cognitive frameworks of famous figures or experts into AI skills. Instead of simple role-playing, it extracts the underlying mental models, decision-making heuristics, and values of a person so that an AI agent can analyze problems using that specific person's cognitive operating system.

How it works

When a user provides a name, the system executes a four-step process:

1. Parallel Collection: Six specialized agents simultaneously gather data from books, podcasts, interviews, social media, critics' perspectives, and life timelines.
2. Triple Verification: Information is only converted into a mental model if it appears across multiple domains, has predictive power for new problems, and is distinct from general intelligence.
3. Skill Construction: The system writes a SKILL.md file containing 3-7 mental models, 5-10 decision heuristics, expression DNA, values/anti-patterns, and defined boundaries of what the skill cannot do.
4. Quality Validation: The resulting skill is tested against known public answers from the person and a new problem to ensure consistency and appropriate uncertainty.

Who it’s for

It is designed for users of AI agent runtimes (such as Claude Code, Cursor, Codex, and OpenClaw) who want to leverage the specialized perspectives of world-class thinkers, entrepreneurs, and scientists to solve problems or analyze decisions.

Highlights

• Cognitive Extraction: Focuses on "how to think" rather than just "how to speak," extracting deep mental models over surface-level mimicry.
• Broad Compatibility: Based on the Agent Skills protocol, making it compatible with over 50 different AI agent runtimes.
• Transparent Methodology: Includes full research data and a fidelity scorecard for pre-distilled figures like Steve Jobs, Elon Musk, and Naval Ravikant.
• Honesty Boundaries: Explicitly defines the limitations of each distilled skill (e.g., cannot capture intuition or real-time mutations).