A recursive feedback framework that enables AI agents to log errors, corrections, and new knowledge to improve future performance.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install feihong-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 feihong-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 technical framework designed to provide AI coding agents with a durable, recursive feedback loop. By systematically capturing command failures, user corrections, and knowledge gaps in structured markdown files, agents can avoid repeating past mistakes and refine their decision-making over time. This skill is a foundational component of the Openclaw Skills ecosystem, ensuring that every session contributes to the long-term intelligence and reliability of the developer workspace.
It bridges the gap between ephemeral chat sessions and permanent project memory by providing a pathway to promote insights into high-level configuration files like CLAUDE.md or SOUL.md. Whether used for individual projects or across a global workspace, this skill transforms agents from static tools into evolving collaborators that learn from every interaction.
Self-Improvement Skill Use Cases
- When a terminal command or operation fails unexpectedly during a development task.
- When a user corrects an agent logic or provides updated project context.
- When an agent discovers a more efficient best practice for a recurring coding pattern.
- When a user requests a capability or feature that does not currently exist in the agent toolkit.
- Before starting a major new task to review past project-specific learnings and errors.
How Self-Improvement Skill Works
- Detection: The agent identifies a learning opportunity, such as a manual user correction, a tool error, or an outdated knowledge gap.
- Logging: The agent appends a structured entry to the appropriate file in the .learnings/ directory using a specific TYPE-YYYYMMDD-XXX ID format.
- Tracking: Issues are tracked with statuses like pending, in_progress, or resolved to manage the technical debt and improvement lifecycle.
- Promotion: Broadly applicable insights are distilled and moved into permanent project memory files like CLAUDE.md or AGENTS.md.
- Extraction: High-value, reusable patterns are converted into independent Openclaw Skills using automated extraction scripts.
Self-Improvement Skill Setup
To get started with this skill in your environment, use the following commands:
Installation via ClawdHub (Recommended):
clawdhub install self-improving-agent
Manual Installation:
git clone https://github.com/peterskoett/self-improving-agent.git ~/.openclaw/skills/self-improving-agent
Initialize Workspace:
mkdir -p ~/.openclaw/workspace/.learnings
Enable Hooks:
openclaw hooks enable self-improvement
Self-Improvement Skill Data Schema & Taxonomy
The skill organizes data into a dedicated .learnings/ directory with three primary log files. Each entry uses a standardized schema to ensure compatibility across Openclaw Skills.
| File | Purpose | Key Metadata Fields |
|---|---|---|
LEARNINGS.md |
Corrections, knowledge gaps, and best practices. | Area, Priority, Pattern-Key, Source |
ERRORS.md |
Command failures and stack traces. | Reproducible, Context, Suggested Fix |
FEATURE_REQUESTS.md |
User-requested agent capabilities. | Complexity, User Context, Frequency |
All entries are timestamped and assigned a unique ID to facilitate grep-based searching and cross-referencing between sessions.
Self-Improvement Skill Advanced Features
- Automatic Hook Integration: Scripts like activator.sh and error-detector.sh provide real-time reminders and automatic logging of terminal failures.
- Inter-Session Communication: Leverage session_send and session_history tools to share learnings between active agents in the Openclaw Skills environment.
- Skill Extraction Pipeline: Automated helper scripts to transform project-specific learnings into reusable, standalone agent skills following the official specification.
- Recurring Pattern Detection: Advanced logic to track the frequency of specific issues and trigger automatic promotion to the system prompt once a recurrence threshold is met.
- Promotion Targets: Seamless integration for updating SOUL.md and TOOLS.md files to refine agent personality and tool-use guidelines.
SKILL.md
Loading
Related Openclaw Skills
ASCII Tool
A specialized utility for inspecting ASCII character codes and performing multi-base conversions between text and numerical formats.
Alias Tool
A streamlined utility for creating, managing, and persisting shell command aliases to accelerate terminal-based development.
Hasbro Intellectual Property Insight
A comprehensive analytical skill designed to research Hasbro's business model, classic IP portfolio, and market dominance in the toy and entertainment sectors.
Backup Tool
A versatile utility for compressing, encrypting, and automating file backups and restorations.
J.M. Smucker Brand Intelligence
A specialized research skill for analyzing the market dominance, historical pivot strategies, and multi-segment portfolio of The J.M. Smucker Company.
Estee Lauder Beauty Market Intelligence
A comprehensive market intelligence skill for analyzing Estee Lauder's prestige beauty portfolio, financial metrics, and global market dominance.
Featured*
