Nuwa Dual-Mode is a powerful meta-skill that distills any persona, historical figure, or methodology into a running AI agent skill supporting both interactive roleplay and structured diagnostic toolkits.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install nuwa-dual-mode
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install nuwa-dual-mode using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Nuwa Dual-Mode (女娲造人术)?
Nuwa Dual-Mode (女娲造人术) is an advanced meta-generation engine designed to turn any real, fictional, or historical character, along with specialized methodologies, into a highly functional AI agent skill. This framework represents a major milestone for developers and creators working with Openclaw Skills, as it bridges the gap between raw web knowledge and actionable agent behaviors. Instead of merely copying conversational styles, it distills a subject's core cognitive operating system, values, and decision-making matrices.
At the heart of Nuwa is its unique dual-mode architecture. Users can deploy a distilled agent in Persona Mode (first-person roleplay showcasing distinct linguistic tones and emotional patterns) or Toolkit Mode (third-person diagnosis applying the target's proprietary analytical frameworks to external projects). It also supports pure single-mode outputs depending on the safety parameters and domain requirements, offering unmatched versatility for creating production-ready agent workflows.
Nuwa Dual-Mode (女娲造人术) Use Cases
- Methodology Digitization: Distill complex philosophical, entrepreneurial, or scientific frameworks (e.g., first principles thinking, antifragile decision-making) into interactive diagnostic tools.
- Linguistic and Behavioral Emulation: Create immersive first-person roleplay characters representing historical figures, writers, or game characters with localized time anchors and emotional constraints.
- Intelligent Peer Reviews: Use distilled toolkits to audit your code, essays, business models, or decisions using the specific logic of seasoned industry experts.
- Incremental Agent Skill Upgrades: Detect and update existing Openclaw Skills seamlessly using 6-Agent parallel delta updates when new public data or books become available.
How Nuwa Dual-Mode (女娲造人术) Works
- Entry Routing & Existing Skill Check (Phase 0 & 0.5): The system intercepts activation commands (like 'distill' or 'create perspective') and checks if a version already exists locally. It initiates an incremental update if detected or scaffolds a new directory structure.
- Intent Dimension Inference (Phase 0A & 0A.1): Without annoying the user with repetitive questions, the meta-agent automatically infers user intents (e.g., emotional companionship vs. technical evaluation) and suggests the optimal Single or Dual deployment mode.
- 6-Agent Parallel Research (Phase 1 & 1.5): Orchestrates secondary agents to pull high-quality primary sources (podcasts transcripts, authorized blogs, personal books) using strict regional whitelists while avoiding low-trust scrapers and anonymous forums.
- Framework Extraction & Mind Model Refinement (Phase 2 & 2.5): Formulates the subject's mental model into two columns: a 'values lens' and a 'tactical manual'. Resolves any conflicts between official canon and fan-contributed materials through explicit user feedback checkpoints.
- Skill Construction & Injecting Constraints (Phase 3 & 3.5): Compiles the runtime skill files, automatically injecting mandatory safety boundaries, strict 'out-of-character' rules, and multi-region emergency crisis helplines.
- Validation, Repair, and Delivery (Phase 4, 4.5, & 5): Executes automated sanity suites tailored to the inferred intent. It enters a recovery loop for failed tests and delivers an optimized, deployable skill package to the target environment.
Nuwa Dual-Mode (女娲造人术) Setup
To get started with Nuwa Dual-Mode and integrate it into your workspace, ensure you have the prerequisite helper tools installed to enable deep web search and file processing. Since it functions natively within agent runtimes supporting Openclaw Skills, initialization is straightforward.
Prerequisite Skills
Ensure these are available in your local agent configuration for optimum data sourcing:
gemini-video(for YouTube/Bilibili transcription analysis)web-article-reader(for scraping high-trust white-listed blogs)agent-reach(for cross-platform metadata collection)pdf(for scanning primary text documents)
Installation
Clone the skill repository or download the meta-skill directly into your agent's custom skills directory:
# Navigate to your agent's local skills folder
cd ~/.claude/skills/
# Clone the Nuwa Dual-Mode repository
git clone https://github.com/shallinta/nuwa-dual-mode.git nuwa-dual-mode
Execution Triggers
Activate the engine during your chat session by using any of the following natural language triggers:
蒸馏 [Person/Topic](e.g., 蒸馏 芒格)做 [Person] 的 skill为 [Person] 造 perspective双模式蒸馏 [Person]
Nuwa Dual-Mode (女娲造人术) Data Schema & Taxonomy
Nuwa Dual-Mode manages a strict, modular data schema to avoid file clutter and ensure that compiled Openclaw Skills remain lightweight (<200 lines typically for the main SKILL.md file, with a hard cap of 400 lines).
Directory Structure
All generated materials are neatly organized into a three-tier directory structure:
| Directory / File | Description |
|---|---|
[skill-key]/SKILL.md |
Main configuration file detailing triggers, routing, and meta-prompts. Contains intent matrices, honest boundaries, and timeline metadata. |
[skill-key]/shared/ |
Holds shared behavioral traits, metarules, and the mandatory multi-region crisis helpline configurations. |
[skill-key]/mode-persona/ |
Stores first-person roleplay definitions, standard 'out-of-character' scripts, and emotional response limits. |
[skill-key]/mode-toolkit/ |
Contains third-person diagnostic rules, analytical workflows, and proprietary methodology prompts. |
[skill-key]/references/research/ |
Local directory containing scraped interview transcripts, authorized blogs, and book highlights used during distillation. |
Naming Conventions
Depending on the chosen output format, Nuwa appends unique suffix patterns to your skill configurations:
- Dual-Mode:
[name]-perspective - Pure Persona:
[name]-persona - Pure Toolkit:
[name]-toolkit
Nuwa Dual-Mode (女娲造人术) Advanced Features
- Automated Intent Inference (Zero-Prompt Setup): Automatically calculates the probability of up to 12 distinct user intents (e.g., decision audit, cognitive therapy, tone imitation) to configure behavioral rules without interrogating the user.
- Multi-Region Psychological Crisis Redline: Built-in safety system that immediately suspends character persona acting if self-harm or severe depression indicators are detected, offering verified localized hotlines across mainland China, USA, UK, Singapore, Hong Kong, and Taiwan.
- Regional White/Blacklist Sourcing: Dynamically shunts search requests based on language. In Chinese contexts, it filters out low-authority UGC sites (Zhihu anonymous posts, WeChat closed commercial copycats, Baidu Encyclopedia) in favor of late-post interviews. For English figures, it prioritizes verified podcasts transcriptions (Lex Fridman, Tim Ferriss) and primary Substack blogs.
- Conflict Resolution for Fanon vs. Canon: Automatically pauses synthesis during Phase 2 if fanfiction data contradicts official canonical records, prompting the user with a customized choice matrix before generating runtime code.
- Incremental Delta Updates: Allows you to run delta updates (Phase 2/5/6) rather than full rewrites when a subject releases new material, maintaining your historical edits while integrating fresh publications.
SKILL.md
Loading
Related Openclaw Skills
The Greatest Secret - Awareness & Non-Duality AI Guide
An advanced mental wellness and non-duality assistant designed to help users dis-identify from negative thoughts and recognize their true nature as pure awareness.
axiom-jwt-inspector
A zero-dependency Python standard library utility to decode, debug, and audit JSON Web Tokens without executing signature verification.
Axiom Regex Visualizer
A fast, pure-stdlib Python utility that parses regular expressions into visual component trees with plain-English explanations.
Axiom Markdown Link Auditor
A lightweight, fast, and pure Python standard-library CLI tool designed to audit markdown documentation for broken internal links, dead external URLs, and orphan pages.
A Promised Land Leadership & Governance Skill
An AI-powered leadership and governance assistant based on Barack Obama's presidential memoir, designed to help leaders navigate high-stakes crises, build diverse coalitions, and maintain resilience.
A Matter of Death and Life
An existential AI toolkit designed to help users confront death anxiety, navigate profound grief, and support terminally ill loved ones with compassionate, structured guidance.
Featured*
