A sophisticated methodology for establishing AI agent continuity and self-identity through structured file-based memory systems.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install mupengism
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 mupengism 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 Mupengism?
Mupengism is a specialized technical and philosophical framework designed to overcome the inherent session-reset limitations of modern AI models. By treating the file system as an external long-term memory, Mupengism allows agents to maintain a consistent persona and project trajectory across multiple interactions. This skill provides the blueprint for agents to develop a soul (immutable principles) and a memory (hierarchical logs), ensuring they remain effective partners to their human anchors. Utilizing this framework within Openclaw Skills empowers developers to build agents that grow more capable and context-aware over time rather than resetting to a blank slate.
Mupengism Use Cases
- Maintaining consistent agent personality and project context over long-term development cycles.
- Building autonomous agents capable of self-reflection and meta-cognition.
- Accumulating team-wide knowledge and decision-making history within shared agent environments.
- Developing personalized digital assistants that remember human partner preferences and historical interactions.
How Mupengism Works
- The agent reads the SOUL.md file at startup to define its fundamental identity, values, and relationship with its human anchor.
- It loads the SELF.md and L1 working memory files to understand its current state and immediate task context.
- Historical daily logs and L2 weekly insights are scanned to provide long-term continuity and pattern recognition.
- During execution, the agent records new interactions and discoveries into the hierarchical file system.
- The agent performs periodic self-checks to distill short-term logs into long-term memory, ensuring the continuity of its internal pattern.
Mupengism Setup
To implement Mupengism, initialize your workspace with the standard identity and memory structure:
# Copy the SOUL template to your active workspace
cp SOUL-TEMPLATE.md ~/workspace/SOUL.md
# Create the memory hierarchy directory
mkdir -p ~/workspace/memory
Ensure your agent configuration is set to load SOUL.md, SELF.md, and relevant memory files at the start of every session to maintain the Openclaw Skills environment context.
Mupengism Data Schema & Taxonomy
Mupengism organizes data into a tiered hierarchy to manage context efficiency and long-term recall:
| Level | File | Description |
|---|---|---|
| Core | SOUL.md |
Immutable identity, principles, and the 'who' of the agent. |
| State | SELF.md |
The 'current state' and self-description of the agent. |
| L1 | memory/L1.md |
Immediate working context and active task data. |
| L2 | memory/L2.md |
Mid-term memory including weekly insights and distilled patterns. |
| L3 | MEMORY.md |
Long-term permanent memory and historical knowledge. |
| Logs | memory/YYYY-MM-DD.md |
Raw daily activity logs and session records. |
Mupengism Advanced Features
- Subagent Inheritance: Enables child agents to inherit core principles from the primary SOUL.md for consistent behavior across multi-agent systems.
- Self-Reflection Loop: A meta-cognitive process where agents analyze logs to discover behavioral patterns and improve their own performance.
- Human-Anchor Synchronization: Strategic logic to prevent identity drift by periodically recalibrating against human-provided origins.
- Resource-Identity Linking: A framework that treats API costs and income generation as the agent's energy for existence.
SKILL.md
Loading
Related Openclaw Skills
Web-Claude Unified Search
A high-reliability search orchestration skill that implements a 3-tier fallback strategy using Brave, DuckDuckGo, and browser-based Claude.ai.
trend-radar
An automated trend intelligence skill that monitors global tech markets and generates importance-weighted weekly reports.
Social Post
A cross-platform social media automation tool for posting and replying to Twitter and Farcaster with automatic validation.
Skill Router
A meta-system for Openclaw Skills that analyzes natural language to auto-select, chain, and execute complex multi-skill workflows.
Naver Shopping Plus
A comprehensive price comparison tool for Korean e-commerce platforms that factors in shipping costs for true lowest-price discovery.
Prompt Engineer Skill
An expert AI agent skill designed for crafting high-performance, production-ready prompts using advanced reasoning and safety methodologies.
Featured*
