A specialized skill that enables AI agents to capture errors, user corrections, and insights into a structured feedback loop for continuous project improvement.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install be-strong
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 be-strong 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 provides a structured framework for AI coding agents to learn from their mistakes and successes in real-time. By logging unexpected failures, direct user corrections, and newly discovered insights into dedicated markdown files, this skill ensures that project-specific knowledge is preserved and shared across sessions. It functions as a persistent memory layer that transforms transient interactions into durable project intelligence.
Built to be a cornerstone of the Openclaw Skills ecosystem, it allows agents to recognize recurring patterns and avoid repeating past errors. The skill encourages a proactive approach to development where the agent doesn't just execute tasks but actively refines its own behavioral guidelines and tool usage patterns over time.
Self-Improvement Agent Skill Use Cases
- Capturing command failures or unexpected API responses for later debugging.
- Documenting user corrections when the agent provides outdated or incorrect information.
- Tracking requested features or missing capabilities that the agent cannot currently perform.
- Identifying and hardening recurring coding patterns to improve codebase stability.
- Promoting temporary learnings into permanent project memory files like CLAUDE.md or AGENTS.md.
How Self-Improvement Agent Skill Works
- Initialization: The skill creates a .learnings directory containing files for learnings, errors, and feature requests.
- Trigger Detection: The agent monitors for specific signals such as non-zero exit codes, user corrections (e.g., "that is wrong"), or identified knowledge gaps.
- Standardized Logging: Insights are recorded using a strict schema including timestamps, priority levels, area tags, and unique IDs (LRN/ERR/FEAT).
- Review and Linking: The agent searches for similar past entries to detect recurring patterns and link related issues.
- Promotion and Extraction: High-value, broadly applicable insights are promoted to project-wide instruction files or extracted into brand new Openclaw Skills.
Self-Improvement Agent Skill Setup
Install the skill via ClawdHub or manual clone into your Openclaw directory:
# Recommended installation
clawdhub install self-improving-agent
# Manual installation
git clone https://github.com/peterskoett/self-improving-agent.git ~/.openclaw/skills/self-improving-agent
# Initialize learning files in your project root
mkdir -p .learnings
To enable automatic reminders in supported agents, you can configure the activator hook:
openclaw hooks enable self-improvement
Self-Improvement Agent Skill Data Schema & Taxonomy
The skill maintains a structured taxonomy within the .learnings/ directory to ensure clarity and machine-readability:
| File | Content Type | Key Metadata Fields |
|---|---|---|
| LEARNINGS.md | Insights, corrections, and best practices | Category, Priority, Area, Pattern-Key, Recurrence-Count |
| ERRORS.md | Command failures and tool exceptions | Error message, Context, Suggested Fix, Reproducible flag |
| FEATURE_REQUESTS.md | User-requested capabilities | Requested Capability, Complexity Estimate, User Context |
All entries follow the naming convention TYPE-YYYYMMDD-XXX (e.g., LRN-20250115-001) to allow for easy referencing and cross-linking.
Self-Improvement Agent Skill Advanced Features
- Automated Skill Extraction: Use built-in scripts to turn verified learnings into standalone, reusable Openclaw Skills.
- Recurring Pattern Detection: Logic to identify issues seen across multiple tasks and escalate their priority automatically.
- Multi-Agent Support: Compatible with Claude Code, Codex, and Openclaw workspace-based prompt injection.
- Promotion Workflow: Rules-based promotion of learnings to SOUL.md, TOOLS.md, or AGENTS.md based on frequency and impact.
- Inter-Session Communication: Tools to send sanitized summaries of learnings between different active agent sessions.
SKILL.md
Loading
Related Openclaw Skills
Novel Writing Assistant
A comprehensive AI-powered assistant designed to automate novel writing, chapter formatting, and story consistency management.
Elsewhere Companion
A digital travel companion that journeys to real-world locations and sends AI-generated postcard updates based on live data.
Conversion Rate Optimization
A comprehensive toolkit for AI agents to perform systematic CRO audits, funnel mapping, and e-commerce experimentation to maximize revenue per visitor.
Nexus Market Researcher
A strategic market research agent designed to perform deep competitive analysis and market sizing across multiple industry verticals.
Capsule Pod Hotel Finder
An AI agent skill designed to locate and book ultra-affordable, tech-forward capsule hotels and pod-style accommodations through real-time CLI data.
Desktop Control
A powerful desktop automation skill for Openclaw Skills that enables remote or local UI control through screenshots, mouse movements, and keyboard interactions.
Featured*
