agent-defibrillator
Watchdog that monitors your AI agent gateway and restarts it when it crashes. Triggers on "install defibrillator", "agent watchdog", "gateway monitor", "auto-restart agent", or "keep agent alive". macOS launchd service with optional Discord notifications.
Why use this skill?
Ensure your OpenClaw AI agent stays online with the Agent Defibrillator. Automatically monitor, detect crashes, and restart your macOS gateway service with Discord alerts.
Install via CLI (Recommended)
clawhub install openclaw/skills/skills/hazy2go/agent-defibrillatorWhat This Skill Does
The Agent Defibrillator is a mission-critical watchdog utility designed for OpenClaw users who require high availability for their AI agent gateways on macOS. Operating as a native launchd service, it acts as a heartbeat monitor, ensuring that your background AI processes remain operational 24/7. It periodically polls your gateway status, detects unresponsive processes or crashes, and triggers an automated recovery sequence. Beyond simple restarts, the skill includes configurable cooldown protections to prevent infinite restart loops, version mismatch detection, and optional webhook-based notifications to keep you informed via Discord if a failure occurs.
Installation
Installation is streamlined for OpenClaw users via the CLI. Run 'clawhub install openclaw/skills/skills/hazy2go/agent-defibrillator' to pull the assets. Once installed, navigate to the local script directory at ~/.openclaw/scripts/defibrillator.sh to configure environment variables. You must ensure your gateway service label matches the 'DEFIB_GATEWAY_LABEL' in the config file. Finally, the service is managed via standard macOS launchctl commands, which provide a robust, OS-level integration for background process maintenance.
Use Cases
- Maintaining continuous uptime for autonomous AI agents that perform long-running tasks.
- Automatically recovering from memory leaks or intermittent segmentation faults in local agent gateways.
- Monitoring background AI infrastructure without manual intervention.
- Getting real-time alerts on production agent crashes via Discord.
Example Prompts
- "Install defibrillator to make sure my agent gateway stays up permanently."
- "Keep agent alive by setting up the watchdog monitor."
- "Auto-restart agent if it crashes, and send me a notification if it does."
Tips & Limitations
This skill is specifically designed for macOS environments and relies on launchd. It will not work on Linux or Windows systems. Ensure you review the bash source code before executing the install script, as it modifies system-level launch agents. Adjust the 'DEFIB_COOLDOWN' setting if you notice your agent frequently failing during boot-up to prevent resource exhaustion. Always verify the status of the watchdog using 'launchctl list' after installation to confirm it is successfully running in the background.
Metadata
Not sure this is the right skill?
Describe what you want to build — we'll match you to the best skill from 16,000+ options.
Find the right skillPaste this into your clawhub.json to enable this plugin.
{
"plugins": {
"official-hazy2go-agent-defibrillator": {
"enabled": true,
"auto_update": true
}
}
}Tags(AI)
Flags: file-write, file-read, code-execution
Related Skills
humanizer
Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Combines Wikipedia's "Signs of AI writing" guide with 2025-2026 forensic detection research. Covers: perplexity/burstiness metrics, Unicode artifacts, 24+ AI vocabulary patterns, structural tells, and surgical humanization techniques. Includes detection benchmarks for GPT-4o, Claude, Gemini models.
humanizer
Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases.
content-research
Research trending topics and generate platform-specific content. Triggers on "research [topic]", "what's new in [topic]", "content for [platform]", "create posts about [topic]". Supports Reddit, X/Twitter, Discord, LinkedIn with multiple content angles per platform.