Skip to content

ngmeyer/personality-dynamics

Repository files navigation

Personality Dynamics

🎭 Dynamic personality evolution for OpenClaw agents. Create rich, personalized AI companions that learn and grow with your interactions.

Features

  • AI-Powered Persona Generation — Generate 4,000+ word character bibles using any LLM
  • Interactive Onboarding — Guided setup without editing files
  • Ongoing Expansion — Add depth through conversation prompts
  • Automatic Evolution — Pattern recognition and weekly insights
  • Mode Switching — Professional / Creative / Casual / Focus personas

Quick Start

# Install from ClawHub
clawhub install personality-dynamics

# Interactive onboarding (recommended for first-time)
npx persona-evolution onboard

# Or generate a rich AI persona
npx persona-evolution generate

Commands

Command Description
init Initialize PERSONA folder with templates
onboard Interactive guided setup
generate AI-powered rich persona generation
expand Add depth via conversation prompts
analyze Analyze session patterns (auto via heartbeat)
report Generate weekly evolution report
mode [set/get/list] Switch personality modes

Setup Guide

Prerequisites

  • OpenClaw installed and configured
  • Node.js (v18+)
  • SOUL.md and MEMORY.md in your workspace

Installation

clawhub install personality-dynamics

First Run

  1. Run onboarding: npx persona-evolution onboard

    • Answer questions about your preferences
    • Sets up PERSONA/ folder with your configuration
  2. Or generate a rich persona: npx persona-evolution generate

    • Creates a detailed character bible prompt
    • Run with your preferred LLM for best results
  3. Enable heartbeat integration (optional):

    • Add heartbeat-check.sh to your heartbeat workflow
    • Auto-analyzes sessions and generates weekly reports

Heartbeat Integration

Add to your HEARTBEAT.md or cron:

bash skills/personality-dynamics/heartbeat-check.sh

This will:

  • Auto-analyze new sessions for personality patterns
  • Generate weekly evolution reports on Sundays

How It Works

Pattern Recognition

The skill analyzes your daily memory files (memory/YYYY-MM-DD.md) for:

  • Communication style preferences
  • Enthusiasm and stress signals
  • Topic engagement patterns
  • Trust level indicators

Mode Switching

Four built-in modes adapt your agent's behavior:

  • Professional — Polished, organized, concise
  • Creative — Enthusiastic, experimental, expressive
  • Casual — Warm, friendly, natural
  • Focus — Ultra-direct, minimal, efficient

Modes auto-detect from context or can be set manually.

Weekly Evolution

Every Sunday, a report is generated with:

  • Session metrics and engagement patterns
  • Emotional context tracking
  • Suggested SOUL.md updates
  • Trust level progression

File Structure

PERSONA/
├── core.md              # Base personality (generated by onboard/init)
├── core-rich.md         # Rich character bible (generated by generate)
├── voice.md             # Voice & communication config
├── interests.md         # Topics & engagement
├── emotional-state.md   # Current emotional context
├── backstory.md         # Character backstory (via expand)
├── learnings.md         # Things taught via expand
├── config.json          # Settings & permissions
├── .current-mode        # Active personality mode
└── evolves/             # Auto-generated insights
    ├── patterns.json    # Detected interaction patterns
    ├── session-*.json   # Per-session analysis
    └── weekly-report.md # Latest weekly report

Environment Variables

Variable Description Default
OPENCLAW_WORKSPACE Workspace root directory Current directory

Privacy

  • All persona data stays local in your PERSONA/ folder
  • .clawhubignore excludes user-specific files from publishing
  • No data is sent externally — analysis runs locally
  • Weekly reports and patterns are stored only in your workspace

License

MIT

About

Agent personality evolution framework (MIT) — generic, sanitized skill for OpenClaw

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors