A self-improving memory framework that enables AI agents to review their work, extract patterns, and compound knowledge daily.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install ai-compound
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 ai-compound 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 Compound Engineering?
Compound Engineering is a specialized framework designed to create a continuous improvement loop for AI agents. It functions by systematically reviewing agent sessions, extracting valuable insights, and updating persistent memory files. This ensures that the agent's knowledge base grows more refined over time, allowing it to anticipate user needs and avoid previously encountered pitfalls. By integrating these Openclaw Skills into your workflow, you transition from using a static model to a dynamic partner that evolves alongside your project.
The core value of this skill lies in its ability to automate the cognitive overhead of documentation. Instead of manually updating project instructions or preference files, the agent assumes responsibility for its own growth. It analyzes decisions, captures recurring patterns, and identifies unfinished tasks, ensuring that the agent you interact with tomorrow is fundamentally more capable and context-aware than the one you used today.
Compound Engineering Use Cases
- Automating nightly post-mortems of development sessions to capture architectural decisions.
- Maintaining a living database of user preferences and project-specific constraints without manual entry.
- Generating granular hourly snapshots to prevent context loss during complex, multi-stage tasks.
- Updating system-level agent instructions (AGENTS.md) based on observed successes or failures.
- Implementing a version-controlled knowledge base that tracks the evolution of an AI's operational expertise.
How Compound Engineering Works
- The AI agent performs its standard tasks, generating a trail of sessions, chats, and decisions throughout the day.
- A scheduled review job, typically running nightly, scans the last 24 hours of activity logs.
- The system extracts specific categories of data: successful patterns, 'gotchas' or failures, user preferences, and pending TODOs.
- Long-term memory files (MEMORY.md) and daily logs are automatically updated with these newly synthesized insights.
- The agent commits these updates to Git, providing a version-controlled history of its own learning process.
- On the following day, the agent reads the updated memory files, starting the session with improved context and refined instructions.
Compound Engineering Setup
Get started with Compound Engineering to enhance your Openclaw Skills experience by following these steps:
# Manually trigger a review of the last 24 hours
npx compound-engineering review
# Initialize hourly memory snapshots
npx compound-engineering snapshot
# Configure the automated nightly review via cron
npx compound-engineering setup-cron
For users integrating with Clawdbot, you can add a nightly review job directly to your configuration to automate the extraction of patterns, mistakes, and user preferences into your global memory files.
Compound Engineering Data Schema & Taxonomy
The skill maintains a structured memory taxonomy to balance high-level wisdom with granular session data:
| Component | File Path | Description |
|---|---|---|
| Long-term Memory | MEMORY.md |
Central repository for successful patterns, gotchas, and user preferences. |
| Daily Logs | memory/YYYY-MM-DD.md |
Specific session summaries, decisions made, and open items for a given day. |
| Agent Instructions | AGENTS.md |
Operational workflows and system prompts updated based on compounded learnings. |
| Hourly Snapshots | memory/YYYY-MM-DD-HH.md |
Optional granular logs for high-frequency task tracking. |
Compound Engineering Advanced Features
- Support for Clawdbot Heartbeat integration to perform real-time memory updates every hour.
- Automated Git integration for versioning memory files with standardized commit messages.
- Customizable extraction logic to focus on specific domains like security audits or API documentation.
- Multi-agent support for synchronizing learnings across a fleet of specialized sub-agents.
- Long-term pattern recognition that identifies workflow inefficiencies over weeks or months of data.
SKILL.md
Loading
Related Openclaw Skills
Component Generator
A CLI-driven AI tool that transforms natural language descriptions into production-ready UI components across multiple frameworks.
Comment Generator
An AI-driven CLI tool that automatically generates meaningful inline comments focusing on logic and intent rather than just syntax.
Codemod Generator
An AI-driven CLI tool that converts natural language descriptions into robust AST-based codemods for safe, large-scale code refactoring.
Cloudflare Gen
Cloudflare Gen is an AI-driven tool that translates natural language prompts into fully configured Cloudflare Workers and edge functions.
Core Vitals Fixer
An AI-powered diagnostic tool that scans your source code to identify and provide actionable fixes for Core Web Vitals performance bottlenecks.
Coverage Boost
An AI-driven tool that analyzes code coverage gaps and generates targeted unit tests to improve software reliability.
METADATA
Featured*
