A recursive learning framework that enables AI coding agents to capture errors, user corrections, and best practices to improve future performance automatically.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install alex-self-improving-agent
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 alex-self-improving-agent 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 Self-Improvement Skill?
The Self-Improvement skill is a powerful framework designed to turn every interaction into a growth opportunity for AI coding agents. By systematically logging unexpected command failures, user corrections, and discovered best practices into structured markdown files, the agent creates a persistent knowledge base that evolves with the project. This ensures that the agent does not repeat the same mistakes and becomes increasingly aligned with specific project conventions and developer preferences.
Built to integrate seamlessly with Openclaw Skills, this tool provides a standardized way to promote transient learnings into permanent project memory. Whether it's a tool-specific 'gotcha' or a complex workflow improvement, the skill ensures that valuable technical insights are never lost between sessions, effectively creating a self-evolving development environment.
Self-Improvement Skill Use Cases
- Automatically logging command or operation failures to prevent recurring bugs.
- Capturing explicit user corrections to refine the agent's behavioral logic.
- Documenting missing feature requests and capability gaps for future implementation.
- Promoting frequently used patterns and best practices to project-wide memory files like CLAUDE.md.
- Sharing technical insights across different agent sessions using inter-session communication tools.
How Self-Improvement Skill Works
- The agent detects a trigger event such as a bash error, a user correction, or a newly discovered best practice.
- Information is captured in a structured format and appended to specific log files in the
.learnings/directory. - The agent generates a unique ID (e.g., LRN-YYYYMMDD-XXX) to track the learning entry and its resolution status.
- Periodic reviews or automated hooks prompt the agent to evaluate pending entries before starting new major tasks.
- High-value, broadly applicable learnings are distilled and promoted to permanent workspace files like SOUL.md or TOOLS.md.
Self-Improvement Skill Setup
To get started with this skill in your Openclaw environment, use the following commands:
# Install via ClawdHub
clawdhub install self-improving-agent
# Create the necessary directory structure
mkdir -p ~/.openclaw/workspace/.learnings
# Optional: Enable the automatic reminder hook
openclaw hooks enable self-improvement
Self-Improvement Skill Data Schema & Taxonomy
The skill organizes data within a .learnings/ directory using three primary markdown logs to maintain a clear taxonomy of project evolution:
| File | Content Type | Primary Metadata Fields |
|---|---|---|
LEARNINGS.md |
Corrections & Best Practices | Category, Priority, Area, Suggested Action |
ERRORS.md |
Command & Tool Failures | Error Output, Context, Reproducibility |
FEATURE_REQUESTS.md |
Requested Capabilities | Complexity Estimate, User Context |
Each entry follows a strict schema including a unique ID, ISO-8601 timestamp, and status tracking (pending, resolved, or promoted).
Self-Improvement Skill Advanced Features
- Multi-agent support for Claude Code, Codex, and GitHub Copilot through customized configuration hooks.
- Automatic skill extraction scripts that transform verified learnings into standalone, reusable Openclaw Skills.
- Inter-session communication tools including
sessions_sendandsessions_listto share knowledge between active sub-agents. - Detection triggers for 'knowledge gaps' that identify when the agent's internal training data is outdated compared to the current project state.
- Hybrid gitignore options to allow teams to choose between private local learnings or shared repository-tracked knowledge.
SKILL.md
Loading
Related Openclaw Skills
Norman Mailer Mental Model Skill
A provocative analytical framework that infuses AI agents with the existential philosophy, power ontology, and aggressive narrative style of Norman Mailer.
For Marketing Editor AI
A cloud-based AI video editor that automates the conversion of raw footage into professional, platform-ready marketing videos in minutes.
Japanese Editor AI
An AI-powered video editing agent that automates Japanese subtitling, styling, and cloud-based rendering for content creators.
Unlimited Editor AI
A cloud-based video editing tool that automates cuts, transitions, and exports without usage limits.
Editor AI HD
A high-performance AI video editing tool that automates cutting, transitions, and 1080p rendering using cloud-based GPU processing.
Algae Feeding Calculator
An automated aquaculture utility that calculates precise algae feeding volumes based on tank size, species strain, and growth stage.
Featured*
