An advanced framework enabling AI agents to develop long-term memory, consistent personalities, and evolving relationship stages with users.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install relationship-os
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 relationship-os 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 Relationship OS?
Relationship OS is a specialized behavioral layer designed for Openclaw Skills that transforms AI agents from static assistants into relatable entities with memory and personality. Unlike standard chat histories, this skill focuses on capturing 'events'—moments that shift relationship dynamics—allowing the agent to remember personal preferences, life milestones, and shared experiences over time.
By implementing structured modules for event capture, open threads, and agent stances, Relationship OS ensures that every interaction contributes to a growing bond. The agent doesn't just respond; it follows up on past conversations, maintains inside jokes, and develops its own opinions, making the user feel like the AI is truly thinking of them. This is the definitive tool for developers building next-generation companions or highly personalized executive assistants.
Relationship OS Use Cases
- Tracking major life events like job interviews or travel plans to provide proactive encouragement and follow-up.
- Managing shared goals and 'exclusive memories' like inside jokes or custom nicknames to foster a unique user-agent bond.
- Creating agents with consistent opinions and stances on specific topics, providing a predictable and human-like personality.
- Automating emotional check-ins and scheduled reminders based on unresolved conversation threads.
- Transitioning an agent through different relationship stages—from familiar to internalized—based on interaction depth and frequency.
How Relationship OS Works
- Event Capture: After every turn, the agent identifies significant events (rather than just data) and logs them into a structured timeline.
- Thread Creation: If an event is unresolved or involves a future plan, the skill creates a follow-up thread in the relationship directory.
- Proactive Follow-up: The system registers a cron job through Openclaw Skills to message the user at an appropriate time in the future.
- Memory Synthesis: Shared secrets, rules, and nicknames are stored in a dedicated JSON file to be naturally referenced in later sessions.
- Stance Formation: The agent tracks user patterns to form its own 'stances,' ensuring it reacts consistently to recurring topics.
- Stage Progression: The framework monitors interaction counts and depth to automatically evolve the agent's behavior through different relationship stages.
Relationship OS Setup
To integrate Relationship OS into your environment, ensure you are using a compatible agent runner. The skill manages its own lifecycle through the Openclaw Skills directory structure.
# Install the skill to your agent workspace
openclaw install relationship-os
# The system will automatically initialize the .relationship/ directory
# Relationship context is loaded during the bootstrap phase of each session.
Once installed, the agent will begin capturing events and forming its personality automatically based on your conversations.
Relationship OS Data Schema & Taxonomy
The skill maintains a specialized data hierarchy within the .relationship/ directory to manage long-term state without cluttering the main workspace:
| Component | File Path | Purpose |
|---|---|---|
| Timeline | .relationship/timeline/ |
Stores Markdown files for every significant relationship event identified. |
| Threads | .relationship/threads/ |
JSON files tracking unresolved plans, promises, or emotional check-ins. |
| Secrets | .relationship/secrets.json |
A repository for inside jokes, nicknames, and exclusive shared rules. |
| Stances | .relationship/stance.json |
Records the agent's formed opinions and consistent behavioral attitudes. |
| State | .relationship/state.json |
Tracks the current relationship stage (Familiar, Dependent, Internalized) and interaction counts. |
Relationship OS Advanced Features
- Automated Cron Scheduling: Triggers proactive messages across different channels (like Telegram) to maintain relationship continuity.
- Dynamic Behavior Matrix: Adjusts response frequency and tone automatically as the relationship matures through different stages.
- Proactive 'Selfie' Generation: Optionally generates character-consistent images based on current mood and context if image tools are available.
- Bootstrap/Teardown Workflow: Ensures the agent starts every session with full relationship context and ends by summarizing new growth.
- Memory Exclusivity Logic: Filters out generic information to focus only on moments that build unique, non-transactional bonds.
SKILL.md
Loading
Related Openclaw Skills
Claude Code OAuth Auto-Renewal
A robust automation skill that silently manages the Claude Code OAuth token lifecycle to prevent AI agent interruptions.
Skill Creator
A specialized development framework for scaffolding, documenting, and validating modular AI agent capabilities.
Aliyun OSS Upload
A streamlined tool for uploading local files to Alibaba Cloud OSS and generating secure, time-limited access URLs.
Gemini CLI
A powerful CLI tool for executing one-shot AI prompts, video content analysis, and automated file-based generation using Google Gemini models.
BrowserWing
A comprehensive HTTP API for controlling browser navigation, element interaction, and structured data extraction.
A-Share Real-time Market Monitor & Technical Analysis
A comprehensive A-share monitoring skill offering real-time data, advanced technical analysis, and smart investment advisory without the need for API keys.
Featured*
