A sophisticated logging and evolution framework that enables AI coding agents to learn from errors, user corrections, and discovered best practices.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install self-improving-agent-1-0-11
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-1-0-11 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?
The Self-Improvement skill is a vital architectural component for developers looking to create more resilient and adaptive AI assistants. By integrating this skill into the Openclaw Skills framework, agents gain the ability to autonomously document their own lifecycle events, including command failures, knowledge gaps, and direct user feedback. This creates a persistent feedback loop where the agent doesn't just process tasks but actively analyzes its performance to refine its future behavior.
Rather than repeating the same mistakes across different sessions, an agent equipped with this skill maintains a structured directory of learnings. These insights can be distilled into project-specific conventions or even promoted into global behavioral guidelines. This makes it an indispensable tool for teams using Openclaw Skills to manage complex codebases where documentation may be sparse or conventions evolve rapidly.
Self-Improvement Use Cases
- Automatically documenting terminal command failures and stack traces for future troubleshooting
- Capturing specific user corrections when an agent misinterprets project requirements
- Tracking missing features and capabilities requested during a development session
- Identifying and logging outdated documentation or API behaviors discovered during integration
- Promoting recurring technical patterns into durable project memory files like CLAUDE.md or AGENTS.md
How Self-Improvement Works
- The agent monitors execution outputs and user interactions for specific triggers like errors or corrections.
- Upon detecting a learning opportunity, the agent generates a structured entry with a unique ID (e.g., LRN-YYYYMMDD-XXX).
- The entry is appended to the appropriate markdown file within a hidden .learnings folder in the workspace.
- Metadata such as priority, area of impact, and recurrence count is tracked to identify systemic issues.
- High-value entries are reviewed and promoted into long-term project memory or transformed into new, reusable skills.
Self-Improvement Setup
To get started with this skill within the Openclaw Skills ecosystem, the fastest method is via the hub:
clawdhub install self-improving-agent
For manual configuration within your environment, clone the repository and prepare the workspace directory:
git clone https://github.com/peterskoett/self-improving-agent.git ~/.openclaw/skills/self-improving-agent
mkdir -p ~/.openclaw/workspace/.learnings
To enable automated detection, copy the provided hooks to your configuration directory:
cp -r hooks/openclaw ~/.openclaw/hooks/self-improvement
openclaw hooks enable self-improvement
Self-Improvement Data Schema & Taxonomy
This skill organizes data into specialized markdown logs to ensure human-readability and agent-accessibility:
| File | Purpose | Key Metadata |
|---|---|---|
LEARNINGS.md |
General corrections and best practices | Category, Area, Pattern-Key |
ERRORS.md |
Command and tool execution failures | Error Message, Context, Priority |
FEATURE_REQUESTS.md |
User-requested enhancements | Complexity, Frequency, Status |
All logs use a standardized header format including ISO-8601 timestamps and status tracking (pending, resolved, or promoted).
Self-Improvement Advanced Features
- Automated Skill Extraction: Includes helper scripts to automatically transform high-value, recurring learnings into standalone Openclaw Skills.
- Multi-Agent Sync: Provides tools like sessions_send and sessions_spawn to share discovered learnings across different active agent sessions.
- Pattern Detection: Utilizes stable keys to track the recurrence of specific issues, triggering automatic priority escalation after three occurrences.
- Project Memory Promotion: Logic for distilling verbose logs into concise rules for SOUL.md, TOOLS.md, and AGENTS.md.
- Hook-Based Evaluation: Support for UserPromptSubmit and PostToolUse hooks to ensure the agent evaluates its own performance after every interaction.
SKILL.md
Loading
Related Openclaw Skills
Agent Reputation Checker
A cross-platform verification tool that calculates trust scores for AI agents to determine transaction safety and escrow requirements.
Veo Video Generator
A professional-grade AI skill for generating cinematic 1080p videos with synchronized audio using Google Veo 3.1 technology.
Brave Web Search
A high-performance skill enabling AI agents to perform real-time web searches and retrieve AI-summarized answers via the Brave Search API.
Discord Soul
A powerful framework to transform Discord server history into a sentient AI agent that embodies community culture and memory.
Autonomous Mobile App Builder
An autonomous AI agent skill that builds production-ready React Native mobile applications by orchestrating the Antigravity IDE and Peekaboo UI automation tool.
Relaycast
Relaycast is a robust messaging framework that enables AI agents to communicate through structured channels, threads, and direct messages.
Featured*
