A sophisticated preference-learning skill for Openclaw Skills that personalizes AI communication style through explicit user feedback.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install chat
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 chat 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 Chat?
The Chat skill is a dynamic adaptation layer designed for Openclaw Skills that enables AI agents to refine their interaction style over time. Unlike static configurations, this skill utilizes a structured learning process to capture explicit user signals regarding tone, formatting, and technical depth. By maintaining an external memory of user-approved preferences, it ensures that every interaction feels tailored to the user's professional requirements without the need for manual prompt engineering.
Chat Use Cases
- Enforcing specific formatting standards like bullet points or concise summaries for code reviews.
- Eliminating repetitive AI filler phrases and unwanted introductory pleasantries.
- Persisting project-specific communication styles across different sessions within Openclaw Skills.
- Managing a blocklist of rejected interaction patterns to prevent the agent from repeating mistakes.
How Chat Works
- The agent monitors the conversation for explicit feedback signals such as "I prefer X" or "Don't do Y."
- Potential preferences are initially logged in an experiments file to track frequency.
- A three-strike confirmation protocol ensures the agent only asks to save a preference after it has been observed three times.
- Once confirmed, the preference is moved to a permanent memory file, while rejected ideas are moved to a blacklist.
- The agent applies these confirmed preferences to future responses, providing transparency by citing its source from the local memory.
Chat Setup
To set up this skill for your Openclaw Skills environment, initialize the dedicated storage directory and tracking files with the following commands:
mkdir -p ~/chat
touch ~/chat/memory.md ~/chat/experiments.md ~/chat/rejected.md
Chat Data Schema & Taxonomy
The Chat skill organizes its learning data within a dedicated directory structure to ensure privacy and persistence:
| File | Purpose | Logic |
|---|---|---|
memory.md |
Stores confirmed preferences | One-line entries, max 50 lines |
experiments.md |
Tracks temporary patterns | Observed 1-2 times |
rejected.md |
Logs explicitly denied styles | Prevents future proposals |
dimensions.md |
Defines preference categories | Internal reference |
Chat Advanced Features
- Three-strike validation: Prevents the agent from learning from outliers or one-off comments.
- Externalized persistence: Memory is stored in localized Markdown files, ensuring preferences survive updates to Openclaw Skills.
- Citation transparency: The agent explicitly mentions when it is applying a learned preference (e.g., "Using bullets from memory.md").
- Conflict resolution: Uses a timestamp-based approach where the most recent explicit instruction overrides previous entries.
SKILL.md
Loading
Related Openclaw Skills
Chat Rooom
A local-first communication and coordination framework that allows multiple AI agents to collaborate, delegate tasks, and maintain shared state within a workspace.
Chainlink Assistant
A technical resource and automation assistant for integrating Chainlink oracle services and managing LINK tokens across decentralized networks.
CGO / Chief Growth Officer
A strategic AI agent skill designed to drive systematic business growth through acquisition loops, retention frameworks, and rapid experimentation.
CDO / Chief Data Officer
A strategic AI skill that acts as a virtual Chief Data Officer to drive data governance, analytics, and AI/ML initiatives.
ChatGPT
A professional framework for mastering ChatGPT through structured prompt packets, intelligent surface routing, and rigorous output quality assurance.
Chef
A professional-grade culinary assistant skill for managing recipes, techniques, and dietary adaptations with technical precision.
METADATA
Featured*
