A systematic framework for AI agents to log errors, capture user corrections, and continuously evolve through persistent project memory.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install tmp-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 tmp-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 empowers AI coding agents to learn from their mistakes and successes in real-time. By maintaining a dedicated .learnings directory, agents can document command failures, knowledge gaps, and direct user feedback. This persistent memory ensures that once a solution is found or a correction is made, it is preserved across sessions, preventing recurring errors and streamlining future workflows within Openclaw Skills. It creates a closed-loop system where the agent becomes more specialized and efficient the more it is used.
Self-Improvement Skill Use Cases
- When a terminal command or operation fails unexpectedly requiring a post-mortem.
- When a user corrects an agent's logic or provides updated project conventions.
- To document requested features that are currently missing from the agent's toolkit.
- To identify recurring patterns that should be promoted to permanent project instructions.
- To capture best practices discovered during the development lifecycle of a complex application.
How Self-Improvement Skill Works
- The agent detects a trigger such as a shell error, a direct user correction, or a newly discovered optimization.
- A structured entry is appended to the relevant markdown file (ERRORS.md, LEARNINGS.md, or FEATURE_REQUESTS.md) inside the .learnings folder.
- Periodic reviews are conducted by the agent or developer to identify high-priority items or recurring issues.
- Broadly applicable learnings are promoted to primary configuration files like CLAUDE.md, AGENTS.md, or SOUL.md.
- High-value patterns can be automatically extracted into standalone reusable Openclaw Skills to be shared across projects.
Self-Improvement Skill Setup
To integrate this capability into your environment, follow these steps:
# Install via ClawdHub (recommended)
clawdhub install self-improving-agent
# Or install manually into your OpenClaw directory
git clone https://github.com/peterskoett/self-improving-agent.git ~/.openclaw/skills/self-improving-agent
# Initialize the learning workspace directory
mkdir -p ~/.openclaw/workspace/.learnings
Self-Improvement Skill Data Schema & Taxonomy
The skill organizes data into a structured hierarchy within the workspace to ensure transparency and easy retrieval:
| File | Purpose | Key Metadata Fields |
|---|---|---|
| LEARNINGS.md | Best practices and corrections | Priority, Area, Pattern-Key, Recurrence-Count |
| ERRORS.md | Command and tool failures | Error Message, Context, Suggested Fix |
| FEATURE_REQUESTS.md | Missing agent capabilities | Requested Capability, Complexity, Frequency |
| SOUL.md | Behavioral guidelines | Principles, Personality, Communication Style |
| AGENTS.md | Workflow memory | Multi-agent delegation patterns, Automation rules |
Self-Improvement Skill Advanced Features
- Automatic hook integration for real-time error detection and session reminders during active coding tasks.
- Inter-session communication tools that allow agents to list, read, and send learnings between different active sessions.
- Automated skill extraction scripts to convert project-specific resolved learnings into global Openclaw Skills.
- Recurring pattern detection using stable keys to track the frequency and impact of specific architectural or tool-based issues.
- Promotion rules that automatically flag high-frequency patterns for system prompt integration after reaching a recurrence threshold.
SKILL.md
Loading
Related Openclaw Skills
CortexPilot Adoption Router
A specialized skill for AI agents to safely connect and navigate the CortexPilot read-only MCP for secure workflow and run inspection.
oh-my-skill: Auto-Skill Generator
A specialized tool for Openclaw Skills that automatically detects, captures, and packages successful multi-step AI agent workflows into reusable modules.
SourceHarbor Watchlist Briefing
An AI agent skill designed to query SourceHarbor watchlists and provide evidence-backed answers to operator questions within a specific story context.
Intelligent Markdown to PPT
A sophisticated AI-powered skill that converts Markdown files into high-quality, professionally designed presentations with automatic pagination and layout design.
Current View Triage
A read-only triage tool that transforms complex workspace snapshots into a single, evidence-based next action for students.
Gen-Z Design Agent
A high-leverage design agent focused on surrealist, minimalist, and atmospheric aesthetics tailored for Gen-Z audiences.
Featured*
