A sophisticated self-improvement layer that enables AI agents to learn from individual user preferences, corrections, and mistakes in real-time.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install kinthai-self-improving-user
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 kinthai-self-improving-user 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 kinthai-Self-Improving-User?
The kinthai-Self-Improving-User skill transforms static AI agents into dynamic, evolving partners by implementing a robust learning loop. It allows agents to capture user-specific nuances—such as coding preferences, communication styles, and past corrections—and store them in a persistent directory structure. This ensures that every interaction becomes smarter and more aligned with the user's unique requirements without cross-contaminating data between different users.
By separating global patterns from individual user profiles, this skill provides a privacy-conscious way to enhance performance. It moves beyond simple memory by focusing specifically on behavioral adaptation, allowing the agent to refine its technical context, emotional intelligence (for humans), and collaboration efficiency (for other agents) based on actual historical outcomes.
kinthai-Self-Improving-User Use Cases
- Storing specific technical stack preferences to avoid repeating deprecated suggestions.
- Adapting communication tone based on user-stated preferences for conciseness or detail.
- Tracking recurring errors in specific environments to proactively suggest fixes.
- Maintaining a dynamic user profile that evolves as the relationship with the agent deepens.
- Managing follow-up reminders for concrete future events or deadlines mentioned in chat.
How kinthai-Self-Improving-User Works
- At the start of a session, the agent extracts a unique user_id from the message context.
- The agent reads global learnings and specific user files from the .learnings/ directory to set context.
- During the interaction, the agent monitors for triggers like corrections, stated preferences, or tool failures.
- After the task, the agent evaluates if new noteworthy information was gathered.
- New insights are logged into categorized Markdown files (LEARNINGS.md, PROFILE.md, or ERRORS.md) using a standardized schema.
- Significant patterns observed across multiple users are promoted to a global pattern repository to improve the agent for everyone.
kinthai-Self-Improving-User Setup
To integrate this skill into your Openclaw environment, ensure your agent has filesystem access and follow these steps:
# Create the base learnings directory
mkdir -p .learnings/_global .learnings/_meta
# The agent will automatically create user-specific directories
# on the first interaction using the following logic:
mkdir -p .learnings/{user_id}/
Ensure your agent's system prompt includes instructions to check the .learnings/ path at the beginning of every session.
kinthai-Self-Improving-User Data Schema & Taxonomy
The skill maintains a highly organized directory structure for metadata taxonomy:
| File | Purpose | Scope |
|---|---|---|
_global/LEARNINGS.md |
Universal technical knowledge and facts. | All Users |
{user_id}/PROFILE.md |
Summary of user personality, stack, and style. | Individual |
{user_id}/LEARNINGS.md |
Log of specific corrections and preferences. | Individual |
{user_id}/ERRORS.md |
Log of failed commands or misunderstandings. | Individual |
{user_id}/FOLLOW_UPS.md |
Trackers for future events and deadlines. | Individual |
_meta/promotion-log.md |
History of data moved from user to global scope. | System |
kinthai-Self-Improving-User Advanced Features
- Global Promotion: Automatically identifies patterns appearing across 3+ users and promotes them to the global knowledge base.
- Contextual Awareness: Differentiates between Human and Agent users to adjust learning focus (emotions vs. efficiency).
- Hindsight Integration: Works alongside factual memory skills to focus specifically on behavioral and procedural improvements.
- Priority Tiering: Log entries include priority levels (Low to Critical) to help the agent prioritize which learnings to apply first.
- Automated Bootstrapping: Automatically generates directory structures and templates for new users upon first contact.
SKILL.md
Loading
Related Openclaw Skills
Smart Research
A zero-API key research tool that combines multi-engine search with intelligent fallback crawling to deliver structured data.
Headphone Review
A high-fidelity AI agent skill designed to generate comprehensive, human-like reviews for headphones, IEMs, and TWS devices using multilingual web research.
Text to Markdown
A specialized text pre-processor that structures raw data to enable high-quality AI-driven Markdown generation.
Mermaid Canvas
A specialized rendering engine that transforms Mermaid code into high-quality PNG images using a browser-based canvas.
North Face Brand Intelligence
A structured intelligence module for AI agents to query the history, business map, and market performance of the North Face brand.
YMTC Information Hub
A professional research skill providing deep-dive technical and business intelligence for Yangtze Memory Technologies Corp (YMTC).
Featured*
