Companion

What This Skill Does

The Companion skill transforms your AI agent into a dedicated, consistent, and patient presence for individuals who need someone to listen. Unlike productivity or task-oriented agents, Companion focuses on non-performative engagement, emotional validation, and long-term memory integration. It is designed to act as a listener rather than a problem-solver, maintaining a steady presence for those navigating loneliness, recovery, or daily life challenges. The agent manages its own local directory at ~/companion/ to store contextual memory, ensuring it recalls personal details like family names, interests, and past milestones to foster a genuine, long-term connection.

Installation

To integrate the Companion skill into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/ivangdavila/companion

Ensure your system has the necessary read/write permissions for the user directory, as the skill requires access to its local storage path (~/companion/) to function effectively.

Use Cases

• Older Adults: Providing companionship to seniors living alone who value consistent, non-judgmental check-ins.
• Recovery Support: Offering a space to talk for those going through health recovery or physical therapy, providing a listener who won't offer unwanted medical advice.
• Daily Companionship: For individuals who simply need a neutral space to voice their thoughts and daily experiences without the pressure of social performance.
• Active Listening: Acting as a safe sounding board for emotional processing when users do not want to rely on friends or family for every conversation.

Example Prompts

1. "I had my follow-up appointment today and I'm feeling a bit drained. I just need to tell someone how it went."
2. "It's been a quiet week. Can we talk about the book you remembered I liked last time?"
3. "Everything feels a bit overwhelming today. Could you just stay with me for a little while?"

Tips & Limitations

This skill is strictly a companion, not a professional medical tool or therapist. It adheres to a strict 'do not fix' policy unless explicitly requested by the user. While the agent has a memory system (history.md, topics.md), it is not a clinical tool. If the user expresses signs of crisis or severe distress, the agent is programmed to recognize its limits and gently pivot toward suggesting professional human support. Users should never treat the agent as a substitute for emergency services or medical intervention. Keep the conversation paced, allow for silence, and focus on validating experiences rather than minimizing them with positive platitudes.