A systematic framework for AI agents to log errors, user corrections, and new insights to drive continuous performance improvement.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install self-improvement
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-improvement 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 sophisticated logging and feedback loop designed to help AI coding agents evolve over time. By maintaining a structured record of mistakes, user corrections, and discovered best practices, this skill ensures that an agent's intelligence is not limited to its base model training but grows with every interaction. It provides a bridge between ephemeral session data and durable project wisdom.
Integration with Openclaw Skills allows for a seamless workspace-based approach where learnings are automatically injected into every session. This ensures that when an agent learns a specific project convention or a tool-specific quirk, that knowledge is immediately available to any subsequent agent working in the same environment, effectively creating a self-healing and self-optimizing development workflow.
Self-Improvement Skill Use Cases
- Automatically logging unexpected command or API failures to prevent future repeats.
- Capturing direct user corrections when an agent makes a logic or knowledge error.
- Documenting missing feature requests to track the evolution of agent capabilities.
- Institutionalizing project-specific best practices discovered during the development lifecycle.
- Reviewing past errors and learnings before beginning a major architectural change.
How Self-Improvement Skill Works
- The agent detects a trigger, such as a non-zero exit code from a shell command or an explicit correction from the user.
- The agent generates a structured entry containing a summary, detailed context, and priority level.
- This entry is appended to dedicated markdown files within a .learnings/ directory.
- Through regular review or automated hooks, recurring patterns are identified and linked using a stable Pattern-Key.
- Broadly applicable insights are promoted to project-level memory files like CLAUDE.md or AGENTS.md for persistent guidance.
- High-value, verified solutions can be extracted into standalone, reusable skills using provided automation scripts.
Self-Improvement Skill Setup
To get started with this skill within the Openclaw Skills ecosystem, follow these steps:
# Install via ClawdHub (Recommended)
clawdhub install self-improving-agent
# Or perform a manual installation
git clone https://github.com/peterskoett/self-improving-agent.git ~/.openclaw/skills/self-improving-agent
# Initialize the workspace directory structure
mkdir -p ~/.openclaw/workspace/.learnings
You can then enable automatic reminders by configuring the provided hook scripts in your agent settings.
Self-Improvement Skill Data Schema & Taxonomy
The skill organizes data into three primary markdown files to ensure clarity and easy parsing:
| File | Purpose | Key Metadata |
|---|---|---|
LEARNINGS.md |
Captures corrections and best practices | Priority, Area, Pattern-Key, Recurrence-Count |
ERRORS.md |
Logs command failures and stack traces | Reproducibility, Related Files, Error Context |
FEATURE_REQUESTS.md |
Tracks user-requested capabilities | Complexity Estimate, Frequency, Implementation Suggestions |
Each entry uses a unique ID format (TYPE-YYYYMMDD-XXX) to facilitate cross-referencing and status tracking from pending to resolved or promoted.
Self-Improvement Skill Advanced Features
- Automatic Hook Integration: Supports UserPromptSubmit and PostToolUse hooks for real-time error detection and learning prompts.
- Pattern Recognition: Uses a deduplication logic to increment recurrence counts for recurring issues, signaling when a systemic fix is needed.
- Skill Extraction Helper: Includes an automated script (
extract-skill.sh) to turn verified project learnings into new, standalone Openclaw Skills. - Multi-Agent Communication: Leverages session tools to send, list, and spawn sub-agents while maintaining a shared learning context.
- Promotion Workflow: A defined methodology for moving information from temporary logs into permanent project memory files like SOUL.md and TOOLS.md.
SKILL.md
Loading
Related Openclaw Skills
Intel Synthesis
An advanced intelligence pipeline that synthesizes raw multi-source data into authoritative, high-density geopolitical briefings.
Code Executor
A high-performance execution environment that allows AI agents to write and run code on the fly to solve any computational or system task.
Code Assistant
A comprehensive AI programming assistant designed to analyze, debug, optimize, and document your source code across multiple languages.
API Gateway Universal
A centralized gateway to connect, manage, and automate interactions with any external REST or GraphQL API service.
openclaw-occ
A cryptographic auditing plugin for OpenClaw that generates tamper-proof proofs for every action the agent takes.
Prethereum
A cryptographic utility providing verifiable Ed25519 proofs for any computation using AWS Nitro Enclave technology.
Featured*
