A robust framework that enables AI agents to log errors, capture user corrections, and systematically evolve through persistent memory files.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install self-improving-agent-3-0-0
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-3-0-0 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 a specialized protocol designed to give AI coding agents a recursive learning loop. By capturing learnings, unexpected errors, and manual corrections in a structured markdown format, agents can avoid repeating past mistakes and bridge knowledge gaps over time. This skill is a core component of the Openclaw Skills ecosystem, enabling agents to transform transient conversation context into durable project memory.
At its core, this skill provides a standardized way to document what went wrong and what was learned. It leverages a dedicated .learnings directory to store logs that can be reviewed, resolved, or promoted to high-level instruction files like CLAUDE.md or AGENTS.md. This ensures that every interaction contributes to a more intelligent and efficient development environment.
Self-Improvement Agent Skill Use Cases
- When an agent command or script operation fails unexpectedly during a task.
- When a user provides a direct correction to an agent's logic or outdated knowledge.
- Tracking feature requests and capability gaps identified during a development session.
- Discovering and documenting better architectural approaches for recurring coding tasks.
- Identifying systemic errors that require promotion to the primary system prompt or project guidelines.
How Self-Improvement Agent Skill Works
- The agent identifies a learning trigger such as an error, a knowledge gap, or a user correction.
- A structured entry is generated and appended to the relevant log file (LEARNINGS.md, ERRORS.md, or FEATURE_REQUESTS.md) within the .learnings directory.
- The entry is assigned a unique ID and metadata including priority, area (frontend, backend, etc.), and a pattern-key for deduplication.
- During subsequent tasks, the agent reviews these logs to ensure previously identified constraints or best practices are followed.
- High-value or recurring entries are promoted to project-level memory files or extracted into standalone Openclaw Skills for use across other projects.
Self-Improvement Agent Skill Setup
To get started with this skill in your environment, use the following commands:
# Recommended: Install via ClawdHub
clawdhub install self-improving-agent
# Manual installation for Openclaw Skills users
git clone https://github.com/peterskoett/self-improving-agent.git ~/.openclaw/skills/self-improving-agent
# Initialize the learning directory structure
mkdir -p ~/.openclaw/workspace/.learnings
For users of Claude Code or Codex, you can also enable automatic reminders by configuring the activator hooks in your local settings.
Self-Improvement Agent Skill Data Schema & Taxonomy
The skill organizes its data into a standardized directory structure to ensure easy parsing by both humans and AI agents:
| File | Purpose | Metadata Captured |
|---|---|---|
LEARNINGS.md |
Corrections and best practices | Priority, Status, Area, Pattern-Key |
ERRORS.md |
Command and tool failures | Error message, Context, Suggested Fix |
FEATURE_REQUESTS.md |
Requested agent capabilities | Complexity, Frequency, Implementation |
Every entry follows a strict ID format: TYPE-YYYYMMDD-XXX (e.g., LRN-20250115-001).
Self-Improvement Agent Skill Advanced Features
- Automatic Skill Extraction: Use built-in scripts to turn a resolved learning into a portable, standalone skill folder.
- Inter-Session Communication: Leverage Openclaw Skills tools like sessions_send and sessions_history to share learnings across different sub-agents.
- Recurring Pattern Detection: Automatically track how many times a specific issue has occurred to trigger a priority bump or systemic fix.
- Hook Integration: Support for UserPromptSubmit and PostToolUse hooks to automatically detect errors and prompt for learning evaluation.
- Promotion Logic: Defined criteria for elevating local learnings into global project memory files like SOUL.md or TOOLS.md.
SKILL.md
Loading
Related Openclaw Skills
SearXNG Search
A privacy-centric metasearch skill that connects AI agents to local or self-hosted SearXNG instances for anonymous, multi-engine web results.
Bubble Integration for AI Agents
A powerful integration for Openclaw Skills that enables AI agents to interact with Bubble no-code applications for data management and workflow automation.
Perigon
A comprehensive financial news and analytics integration that allows AI agents to track market-moving news and organizational data.
Botmaker
An automated interface for the Botmaker platform to manage chatbots, automate customer interactions, and handle messaging data.
Resend CLI Skill
A professional-grade email infrastructure skill for automating transactional emails, domain management, and marketing workflows via the Resend API.
Admirarr
A unified CLI and AI agent skill for managing, monitoring, and troubleshooting self-hosted media stacks including Jellyfin, Plex, and the *Arr suite.
Featured*
