Nuwa.skill

What is Nuwa’s skill?

Nuwa.skill is an open-source Skill developed by independent developer Hua Shu. It’s not simply an AI role-playing tool, but a “mind distillation” system. Input any name, and six agents collect publicly available data in parallel. Through triple verification, the system extracts the user’s cognitive operating system, ultimately generating a conversational AI Skill that includes a mental model, decision-making heuristics, expressive DNA, and honesty boundaries. Unlike ordinary role-playing, Nuwa doesn’t aim to “resurrect” the user; instead, it uses your questions to run the user’s thought process framework. The information sources and extraction logic for each Skill are completely open and transparent.

The main functions of Nuwa.skill

• Six parallel data collection channels : Six research agents are deployed to simultaneously capture the target person’s works, in-depth interviews, social media activity, critical perspectives, key decision records, and life timeline, comprehensively covering publicly available information sources.
• Triple verification and refinement : The collected viewpoints are rigorously screened, and only those that simultaneously meet the three criteria of “cross-domain reproducibility, problem-generating power, and cognitive exclusivity” will be included as formal mental models.
• Automatic Skill Generation : The system outputs a structured cognitive operating system, including 3-7 mental models, 5-10 decision heuristics, expressive DNA, core values ​​and anti-patterns, and a clear statement of honesty boundaries.
• Quality verification closed loop : Consistency of direction is tested through known issues, boundaries are inferred through edge issues, and style is tested to ensure expression matching, preventing overconfidence or generalization in skill.
• Thematic Skill Distillation : It can synthesize the methodologies of multiple experts for a specific field (such as X Operations) to generate a domain-level cognitive framework.
• Transparent auditing across the entire process : The directory of each Skill references/fully preserves both primary and secondary information sources, research documents, and extraction logic to ensure that the generation process is traceable and verifiable.
• Ecological quality evolution : The “Darwin.skill” assessment system is used to score and automatically optimize all skills in batches from 8 dimensions, continuously improving the overall quality level.

How to use Nuwa.skill

• To install Claude Code : First, ensure that Anthropic’s command-line programming assistant, Claude Code, or skills.sha third-party environment compatible with the protocol (such as OpenClaw) is installed locally.
• One-click deployment of Nuwa : Execute in the terminal npx skills add alchaincyf/nuwa-skillto add Nuwa.skill to Claude Code’s skill library.
• Initiate a distillation command : In the Claude Code dialog box, enter “Distill an XXX” or “Create a skill about XXX”, where XXX is replaced with the target person’s name or domain topic.
• Fully automated operation : Nuwa automatically schedules 6 agents to collect publicly available data in parallel, refine, verify, and generate the final output SKILL.mdfile, with the entire process taking approximately 2-3 hours without human intervention.
• Direct Skill Invocation : Once generated, the skill can be invoked directly in the dialogue to answer your questions using the target person’s cognitive framework, such as “Evaluate this product from Steve Jobs’ perspective.”
• Quality review and optimization (optional): Use the accompanying “Darwin.skill” tool to score the newly generated Skill across 8 dimensions, manually adjust or re-distill based on feedback, and continuously improve output quality.
• Transparent audit traceability : Each Skill’s references/directory contains complete research documents and information sources, allowing for retrospective verification of the generation logic and data basis at any time.

Nuwa.skill’s core advantages

• It’s not about repeating quotes, but about understanding the thought process : capturing “how they think,” not “what they said.”
• It’s not a resurrection technique, but a cognitive mirror : using your questions to operate within a person’s framework, rather than pretending that person has descended.
• Built-in honesty boundaries : Each skill explicitly states “what I cannot do”, which is itself part of the credibility.
• Transparent and auditable : Each Skill’s references/directory contains complete research documents, and the information sources and extraction logic are all publicly available.
• Ecosystem Continuous Evolution : The “Darwin.skill” suite uses an 8-dimensional evaluation system to batch optimize the quality of all skills.

Key information and usage requirements of Nuwa.skill

• Project nature : Open-source AI mind distillation system, not a role-playing tool, designed to generate a working cognitive operating system.
• Developer : Hua Shu, an independent developer (representative work: Cat Fill Light), whose project gained tens of thousands of followers within a week.
• Core mechanism : Six agents collect publicly available data in parallel, and extract a mental model after undergoing triple verification of “cross-domain reproducibility, generative power, and exclusivity”.
• Outputs : Standardization SKILL.md, including 3-7 mental models, 5-10 decision heuristics, expressive DNA, values ​​and anti-patterns, and honesty boundaries.
• The fundamental difference from role-playing : it doesn’t mimic tone or repeat quotes, but rather uses the user’s questions to run through the target character’s thought process framework to make inferences.
• Quality Support : Built-in “Darwin.skill” 8-dimensional evaluation system, supporting batch scoring and continuous optimization of generated skills.
• Environment Requirements : Claude Code or a compatible skills.shthird-party environment (such as OpenClaw) must be pre-installed.
• Installation method : Execute via terminal npx skills add alchaincyf/nuwa-skillto deploy to the skill library with one click.
• Startup command : Enter “distill XXX” or “create a skill related to XXX” in Claude Code to trigger the fully automated process.
• Runtime : A single distillation process takes approximately 2-3 hours, during which 6 agents automatically complete the collection, extraction, verification, and generation without human intervention.
• Material Requirements : The target person must have sufficient publicly available materials (books, speeches, social media posts, etc.). Insufficient materials will lead to a decline in distillation quality.
• Privacy Boundaries : Distillation of private data (emails, Slack, internal documents) is strictly prohibited. Explicit informed consent from the individual is required for any non-public information involved.
• Time Limitation : Skill is a static snapshot at the time of the research. If the core ideas of the character change significantly, the distillation process must be manually rerun to update.
• Validation Recommendation : After generation, it is recommended to use “Darwin.skill” for 8-dimensional scoring, and to optimize or supplement the materials for low-scoring dimensions.