A sophisticated logging and feedback framework that enables AI coding agents to autonomously capture learnings and fix recurring errors.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install self-improving-agent-local
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 self-improving-agent-local 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 Agent Skill?
The Self-Improvement skill is designed to transform transient AI interactions into a permanent knowledge base for continuous optimization. By providing a structured system for logging unexpected command failures, user corrections, and discovered best practices, it ensures that an agent never makes the same mistake twice. As a core component of the Openclaw Skills ecosystem, it bridges the gap between individual sessions and long-term project memory.
This skill empowers developers to build more resilient AI workflows by identifying knowledge gaps and workflow bottlenecks in real-time. Whether it is tracking a minor UI correction or a critical infrastructure failure, the skill synthesizes these events into actionable data that can be promoted to global project instructions or extracted into entirely new automated capabilities.
Self-Improvement Agent Skill Use Cases
- Capturing unexpected terminal command or script failures for later debugging.
- Documenting specific user corrections to align the agent with project-specific coding styles.
- Tracking requests for missing features to inform future development roadmaps.
- Identifying outdated documentation or knowledge gaps in the agent's internal model.
- Discovering and standardizing better architectural approaches for recurring development tasks.
How Self-Improvement Agent Skill Works
- Detection: The agent monitors for specific triggers like command exit codes, explicit user corrections, or discovered optimizations.
- Structured Logging: Upon detection, the agent appends a detailed entry to the local .learnings/ directory, categorizing it as an error, learning, or feature request.
- Contextual Awareness: During active sessions, the agent references previous logs to apply past fixes to current tasks.
- Knowledge Promotion: High-value insights are promoted to project-wide files like CLAUDE.md, AGENTS.md, or SOUL.md to become permanent guidelines.
- Skill Extraction: Recurring patterns can be automatically packaged into new standalone Openclaw Skills using built-in helper scripts.
Self-Improvement Agent Skill Setup
To integrate this skill into your environment, use the ClawdHub registry for the most streamlined experience with Openclaw Skills:
clawdhub install self-improving-agent
For manual installation within your workspace:
git clone https://github.com/peterskoett/self-improving-agent.git ~/.openclaw/skills/self-improving-agent
mkdir -p ~/.openclaw/workspace/.learnings
To enable automated reminders and error detection, activate the provided hooks:
openclaw hooks enable self-improvement
Self-Improvement Agent Skill Data Schema & Taxonomy
The skill manages information through a standardized directory structure and metadata taxonomy within the .learnings/ folder.
| File | Category | Primary Metadata Fields |
|---|---|---|
| LEARNINGS.md | Corrections & Best Practices | ID, Priority, Area, Pattern-Key, Status |
| ERRORS.md | System & Tool Failures | ID, Error Message, Context, Suggested Fix |
| FEATURE_REQUESTS.md | Missing Capabilities | ID, Complexity, Frequency, User Context |
All entries utilize a unique ID format (TYPE-YYYYMMDD-XXX) to allow for easy cross-referencing and pattern tracking across the codebase.
Self-Improvement Agent Skill Advanced Features
- Automated Skill Extraction: Seamlessly turn verified fixes into reusable Openclaw Skills with the included extraction helper script.
- Inter-Session Messaging: Use built-in tools like sessions_send and sessions_history to share learnings across different active agent sessions.
- Multi-Agent Compatibility: Native support for Claude Code, Codex, and GitHub Copilot through customized prompt injection and setting configurations.
- Recurring Pattern Detection: Leverages Pattern-Keys to automatically detect and escalate issues that occur across multiple distinct tasks.
- Real-time Hooks: Optional activator and error-detector scripts that provide immediate feedback loops during the development process.
SKILL.md
Loading
Related Openclaw Skills
Agent Audit Scanner
A robust security auditing tool designed to detect vulnerabilities, credential leaks, and unsafe code within Openclaw Skills before they are executed.
Omnis Venture Intelligence
A machine-first venture intelligence skill for identifying and scoring high-signal startup opportunities via read-first API workflows.
Session Recall
A context-recovery tool that allows AI agents to search and retrieve information from past session transcripts.
Soul Blocks
A powerful tool for minting and managing permanent, on-chain identities and behavioral traits for AI agents on the Base network.
YZTurboWebAndroid
A high-performance Android WebView container SDK featuring preloading, instance reuse, and seamless JS-Native bidirectional communication.
News Daily
An automated RSS news aggregator that fetches global, tech, and AI news and delivers them as formatted cards to Feishu via Openclaw Skills.
Featured*
