A Zen-inspired self-inquiry engine that generates mindful huatou prompts from user emotions and thoughts using static mapping and dynamic LLMs.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install now-huatou-engine
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 now-huatou-engine 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 now-huatou-engine?
The now-huatou-engine is a unique mental wellness and self-reflection skill designed for AI agents. By integrating this module with Openclaw Skills, developers can build agents capable of guiding users through emotional discomfort, anxiety, or cognitive loops using Zen-inspired huatou inquiries. Rather than offering superficial comfort or unsolicited advice, the engine challenges the ego directly, prompting users to observe their attachments.
Starting from a static template system in its MVP phase, the skill scales up to dynamic generation using LLM models. It parses the user's emotional scenes (such as overthinking, decision paralysis, or anxiety) and structures mindful questions that help individuals shift perspective, fostering resilience and mental clarity through structural Openclaw Skills integrations.
now-huatou-engine Use Cases
- Anxiety and Stress Management: When a user expresses work-related stress or anxiety, the engine generates a prompt like "Who is anxious?" to pivot them toward self-observation.
- Overthinking & Decision Paralysis: Dissects complex thought loops and comparison traps, returning short, focused questions to ground the user.
- Emotional Regulation: Deconstructs active anger, sadness, or shame by prompting reflection on the transient nature of emotional states.
- Mindfulness & Meditation Apps: Serves as a backend logic provider for apps running on Openclaw Skills to offer real-time Zen prompts.
How now-huatou-engine Works
- Input Scene Identification: The engine accepts user expressions (such as "my boss criticized me today") or pre-selected scene IDs (like anxiety, overthinking, or free-form).
- Tag & Attachment Mapping: It maps scenes to specific tags. In dynamic V2, it identifies deep-seated mental attachments (e.g., self-attachment, time-attachment, emotional attachment).
- Huatou Generation: In static mode, it selects a template from a JSON library and replaces placeholders like
{emotion}or{time_word}. In dynamic mode, it calls qclaw/modelroute to invoke an LLM Zen persona. - Safety Filter Check: Scans inputs and generated text for high-risk topics like self-harm or deep psychological crises. If found, it bypasses generation to display emergency helpline support.
- Structured Output Response: Delivers a JSON package containing analyzed metadata and a single Zen-style query under 15 characters.
now-huatou-engine Setup
To deploy now-huatou-engine within your application utilizing Openclaw Skills, follow these configuration steps:
1. Initialize Static Templates (V1)
Create the file data/huatou-templates-v1.0.json with your base templates. Ensure it is accessible by your local skill runtimes.
2. Configure Dynamic LLM Routing (V2)
Define your qclaw routing configurations. Make sure your local environment variables have access to your language model provider:
# Set your model router endpoint
export QCLAW_MODEL_ROUTE="https://api.yourdomain.com/v1"
export LLM_API_KEY="your-api-key-here"
3. Register the Skill
Register the engine within your workspace agent config so it can hook into active conversational frameworks.
now-huatou-engine Data Schema & Taxonomy
The engine processes data through two primary schemas, illustrating how Openclaw Skills structure internal logic:
V1 Static Template Mapping (scene_tags Mapping)
| Scene ID | Tag Mappings | Default Override Example |
|---|---|---|
anxiety |
["anxiety", "fear"] |
{emotion} -> 焦虑 (Anxiety) |
overthinking |
["overthinking"] |
{emotion} -> 想 (Thinking) |
decision |
["decision", "confusion", "comparison"] |
{emotion} -> 困惑 (Confusion) |
emotion |
["anger", "sadness", "judgment", "shame", "pain"] |
{emotion} -> 生气 (Anger), 难过 (Sadness) |
free |
[] (Random pick from all) |
N/A |
V2 Dynamic JSON Output Schema
When processing dynamic prompts via the LLM router, the system returns structured JSON data containing:
{
"scene": "The identified mental barrier or scenario",
"attachment": "Identified target of attachment (e.g., ego, expectations)",
"huatou": "The generated Zen question (under 15 characters)"
}
now-huatou-engine Advanced Features
- Dynamic Attachment Recognition: Analyzes the core psychological blocks (self-attachment, time-attachment, dharma-attachment) to target the root of user distress.
- Hardened Crisis Safety Guardrails: Includes a keyword-driven safety filter (monitoring words like suicide or despair) that instantly redirects to helpline configurations instead of generating questions.
- Zero-Advice Design: Adheres strictly to a zen-like philosophy that avoids leading questions, commands, or patronizing emotional comforts, empowering users to find answers internally.
- Flexible Deployment Models: Supports offline-first static templates as well as dynamic LLM integrations depending on resource availability within Openclaw Skills ecosystem environments.
SKILL.md
Loading
Related Openclaw Skills
now-practice
An elegant, lightweight 60-second mindfulness practice engine designed to guide users through anxiety, overthinking, and emotional storms in four simple steps.
Mars Text Translator
A fun and deterministic text converter that transforms standard Chinese into 2000s-era internet 'Mars text' while preserving readability.
dargue-flag
A command-line integrated video search and streaming assistant designed to help users browse, filter, and view detailed video content safely and efficiently.
A-Share Quantitative Finance Toolkit
A lightweight, zero-dependency quantitative analysis toolkit for Chinese A-shares powered by free, real-time Tencent Market Data APIs.
Recruitment Fullstack V4
Recruitment Fullstack V4 is an advanced, committee-level AI hiring framework that automates and structures the entire interview lifecycle from intake checking to calibration and final decision-making.
IP Character Design to 3D Print Pipeline
An automated end-to-end AI pipeline that converts character prompts into production-ready, manifold STL files for 3D printing.
Featured*
