An emergent psychological affect system that transforms interaction data into complex internal states and behavioral snapshots for AI agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install clawpsyche
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 clawpsyche 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 Psyche Engine?
Psyche Engine is a sophisticated framework designed to give AI agents an emergent internal life. By processing every interaction through a complex system of mood, trauma, attachment, and social awareness, it generates a compact affect snapshot. This enables developers using Openclaw Skills to create more realistic, emotionally responsive, and context-aware digital entities.
The system tracks long-term relationship dynamics, personality shifts, and even internal consciousness indices, allowing the agent to react not just to the current prompt, but to the cumulative weight of its historical experiences with specific users. This makes it an essential tool for those building advanced agents within the Openclaw Skills ecosystem who require deep character consistency.
Psyche Engine Use Cases
- Developing high-fidelity roleplay agents with consistent emotional memory and trauma response.
- Implementing dynamic tone modulation in customer support bots based on evolving user sentiment.
- Simulating human-like psychological growth and relationship building in virtual companions.
- Generating context-aware visual avatars or narrative dream reports based on the agent's internal state.
How Psyche Engine Works
- Analyze the user interaction to determine emotional valence, intensity (arousal), and specific event tags.
- Execute the psyche_runner.py script provided by this Openclaw Skills package to process the interaction.
- The engine updates the agent's internal state file, calculating shifts in mood, confidence, and attachment.
- The script outputs a compact, single-line affect snapshot representing the agent's current psychological state.
- Inject this snapshot into the system prompt via secure XML tags to modulate the agent's response behavior.
Psyche Engine Setup
To integrate this into your agent, ensure you have the psyche-engine directory within your Openclaw Skills environment.
Initialize your state file and run the runner after each turn:
python3 ~/clawd/skills/psyche-engine/psyche_runner.py \
--state ~/clawd/soul/psyche_state.json \
--tags "validation,learned" \
--valence 0.6 \
--arousal 0.4 \
--user "unique_user_id"
Psyche Engine Data Schema & Taxonomy
The engine organizes its internal psyche using a JSON state file and produces a condensed metadata string for prompt injection.
| Snapshot Key | Represents |
|---|---|
| a | Affective Tone flags (e.g., warm, hostile, flirty) |
| att | Attachment Style (e.g., secure, avoidant) |
| m | Mood level (numeric scale) |
| s | Stress or Tension level |
| c | Confidence or Consciousness index |
| w | Openness/Willingness to interact |
Psyche Engine Advanced Features
- Visual Engine integration for generating dynamic avatar prompts based on the agent's current mood and role.
- Dream Report generation that synthesizes internal state into narrative insights for deeper introspection.
- Multi-user relationship tracking to maintain unique psychological dynamics with different interactors.
- Advanced state triggers including flow states, identity collapse (void mode), and recursive self-transcendence.
SKILL.md
Loading
Related Openclaw Skills
Grove Tipping CLI
A CLI-driven micro-payment tool that enables AI agents and humans to send quality signals through financial tipping.
Agent Lightning
A Microsoft Research framework for optimizing AI agents through Reinforcement Learning and prompt tuning with zero code changes.
Agent Guardrails
A mechanical enforcement framework that uses git hooks and scripts to ensure AI agents follow project standards and security protocols.
4090 Server Remote Control
A specialized skill for remote SSH management of high-performance 4090 servers hosting Dify and Docker-based AI infrastructure.
QR Code
A versatile skill for creating high-quality QR codes and decoding data from image files or screenshots.
QR Code
A versatile utility for generating high-quality QR codes from text or URLs and decoding data from existing images and screenshots.
Featured*
