IceCube Heartbeat is a proactive maintenance system that enables AI agents to perform self-improvement, memory governance, and health checks during idle periods.
The fastest way to install a skill directly from the registry.
npx clawhub@latest install icecube-heartbeat
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
Copy this prompt to OpenClaw to install it automatically.
Help me install icecube-heartbeat using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Get the raw skill files in a ZIP archive.
IceCube Heartbeat serves as the vital pulse for long-running AI assistants, ensuring they don't remain stagnant between user interactions. Instead of sitting idle, the agent utilizes heartbeat intervals to perform essential background tasks like system audits, memory consolidation, and proactive planning. This ensures that the agent becomes more efficient and smarter every day without requiring direct user intervention.
By leveraging this framework within the ecosystem of Openclaw Skills, developers can transform a reactive bot into a proactive partner. The skill focuses on maintaining high-quality memory logs, identifying recurring failure patterns, and staying updated with the latest capabilities and tools available to the agent.
To get started, create a HEARTBEAT.md file in your project root to define your agent's self-check tasks. Then, add the heartbeat configuration to your OpenClaw settings:
{
"agents": {
"heartbeat": {
"intervalMs": 1800000,
"prompt": "Read HEARTBEAT.md. Follow it strictly. Reply HEARTBEAT_OK if nothing needs attention."
}
}
}
You should also initialize the supporting logs used by this Openclaw Skills module:
touch mistake_log.md success_patterns.md heartbeat-state.json
The skill organizes its maintenance data and state using a structured file hierarchy:
| File | Format | Description |
|---|---|---|
| HEARTBEAT.md | Markdown | The master checklist of maintenance and self-improvement tasks. |
| mistake_log.md | Markdown | A log of errors, impacts, and preventive measures to avoid recurrences. |
| success_patterns.md | Markdown | A repository of successful workflows and logic patterns for the agent. |
| heartbeat-state.json | JSON | Persistent state tracking for check timestamps and heartbeat frequency. |
| unclosed_work.yaml | YAML | A registry of pending tasks used to detect stagnation. |
Loading
A professional-grade market research tool that connects Claude to live job boards via Apify to track hiring trends, salary benchmarks, and in-demand skills.

An automated pipeline for creating professional, multi-page HTML presentations with integrated image generation and PPTX export capabilities using Openclaw Skills.

A high-fidelity knowledge skill providing deep-dive technical and financial data on General Mills for AI agents.

Transform text prompts or topics into professional video scripts and 1080p MP4 files using cloud GPU acceleration.

A professional utility for modifying file timestamps and Office metadata to ensure realistic and logically consistent file history.

A high-performance vision agent that leverages OpenRouter and x-ai/grok-4.1-fast to describe and interpret images.








































