Self-Improving Plus enables AI coding agents to autonomously track errors, capture user corrections, and distill recurring patterns into fully reusable custom skills.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install self-improving-plus
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-plus 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-Improving Plus?
Self-Improving Plus is a cutting-edge, autonomous learning framework designed to elevate your AI coding experience. Built specifically for Openclaw Skills, this framework allows your AI agents to continuously self-correct by logging errors, user corrections, and feature requests. Instead of making the same mistake twice, the agent monitors its own output, identifies friction points, and documents them directly inside a localized repository.
By leveraging the power of Openclaw Skills, Self-Improving Plus transforms raw developer logs into structured, actionable intelligence. It connects seamlessly with GitHub high-star project databases, allowing the agent to research external best practices and dynamically close knowledge gaps. This creates a highly collaborative, self-healing environment where the agent grows smarter with every execution.
Self-Improving Plus Use Cases
- Reducing repetitive agent mistakes: Automatically logs when commands fail or when you issue corrections, ensuring the agent adapts instantly.
- Designing reusable custom tools: Easily promotes resolved bugs and custom logic patterns into standalone Openclaw Skills.
- Researching external best practices: Automatically queries GitHub's most successful projects to find elegant, high-performance solutions to complex tasks.
- Maintaining long-term project context: Keeps a clean, version-controlled audit trail of lessons learned, which stays with your project across coding sessions.
How Self-Improving Plus Works
- Dynamic Incident Logging: When a command fails or you provide a direct correction, the agent instantly catches the event and logs a detailed markdown entry in the local
.learnings/directory. - Pattern Analysis & Refinement: The agent reviews recent entries during natural breakpoints to find common categories, tags, or recurrence trends.
- Knowledge Base Research: If a recurring knowledge gap is detected, the agent queries curated high-star GitHub repositories to ingest top-tier open-source solutions.
- Skill Distillation: If a pattern matches the extraction criteria, the agent packages it into a standard directory layout with its own metadata.
- Deployment & Scaling: The newly created tool is integrated back into the local environment or shared with the developer community as part of your custom Openclaw Skills collection.
Self-Improving Plus Setup
To activate the Self-Improving Plus skill, ensure your local workspace directory is prepared to capture learnings. Create the required storage directory in your repository root:
mkdir -p .learnings
Configure your AI agent instructions to point to this directory and authorize it to monitor, write, and promote entries. Once initialized, the agent will seamlessly start integrating with Openclaw Skills to optimize its performance.
Self-Improving Plus Data Schema & Taxonomy
Self-Improving Plus stores its operational state and findings in a standardized format inside your repository. This makes it highly portable and compatible with other Openclaw Skills.
Directory Layout
.learnings/ERRORS.md: A structured list of command and execution failures..learnings/LEARNINGS.md: Direct user corrections and best practice discoveries..learnings/FEATURE_REQUESTS.md: Unimplemented user features logged for future development.skills/: Target folder where distilled skills are automatically written.
Learning Entry Structure
| Field | Type | Description |
|---|---|---|
| ID | String | Standardized identifier (e.g., LRN-YYYYMMDD-XXX) |
| Category | String | Category tag such as correction, best_practice, or pattern |
| Priority | String | Priority scale: low, medium, high, critical |
| Area | String | Codebase impact area (e.g., backend, infra, tests) |
| Pattern-Key | String | Optional hook for grouping recurring events (e.g., simplify.dead_code) |
Self-Improving Plus Advanced Features
- Automatic Skill Distillation: Evaluates logs against four robust criteria (recurring, verified, non-obvious, and broadly applicable) to auto-extract raw code into reusable skills.
- GitHub High-Star Integration: Deep-dives into top-tier GitHub codebases to extract refined patterns, keeping your local Openclaw Skills aligned with industry standards.
- Structured Meta-Tagging: Utilizes pattern-keys, recurrence counts, and timeline tracking to build a rich telemetry dataset of your workspace development patterns.
- Natural Breakpoint Reviews: Prompts periodic reviews before major tasks or after feature completion, ensuring your project instructions stay updated and precise.
SKILL.md
Loading
Related Openclaw Skills
Code Review Plus
An advanced code review assistant with specialized AI/ML checks, robust security scanning, and multi-language standard compliance.
Browser Auto Plus
An advanced browser automation skill featuring multi-browser automatic failover, exponential backoff error recovery, screenshot verification, and built-in anti-detection web scraping capabilities.
Data Analysis Plus
An enhanced data analysis kit providing comprehensive Python and R code templates, statistical methods, visualization guides, and automated reporting formats.
Web Scrape Plus
An enhanced data extraction tool for AI agents featuring stealth browser automation, proxy routing, rate limiting, and robust structural data validation.
Obstetrics and Gynecology PMPH 10th Edition Clinical Skills
A structured registry of 187 authoritative obstetrics and gynecology clinical skills from the PMPH 10th edition textbook for context-aware AI medical agents.
reading-squad
The reading-squad skill simulates a structured, multi-expert discussion panel inside any AI chat system to extract deep book insights.
Featured*
