A control-plane safety skill designed to prevent agent deadlocks and provide emergency recovery commands for long-running workloads.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install error-guard
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 error-guard 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 error-guard?
error-guard serves as a critical safety layer for Openclaw Skills, acting as a non-blocking control plane that monitors system health. Unlike standard skills that rely on LLM reasoning, this tool provides deterministic commands to manage task states, inspect registries, and recover from unresponsive states without requiring a full container restart. It is an essential utility for developers building complex, long-lived, or high-risk agentic workflows where reliability is paramount.
The skill is built on fail-safe principles, ensuring that the main agent never blocks on external I/O or model calls. By maintaining a minimal state and tracking only task metadata, it provides a robust last line of defense for your automation environment.
error-guard Use Cases
- Managing long-running background tasks like web scraping or large-scale data processing.
- Recovering from agent freezes during sub-agent orchestration or recursive loops.
- Monitoring the health of background monitors such as Moltbook or PNR checks.
- Executing emergency stops to clear pending message queues and reset task registries.
How error-guard Works
- The skill acts as an event-driven listener that tracks task metadata without blocking the main agent execution.
- Users or sub-agents invoke non-LLM commands like /status to inspect the current state of active tasks and heartbeats.
- If a deadlock or freeze is detected, the /flush command kills active processes and clears the message queue.
- The /recover sequence performs a reset of the control-plane state and reloads configuration without killing the container environment.
error-guard Setup
To install this safety-critical skill within your environment, ensure you are running a compatible version of the framework and add the skill via your configuration file or CLI. Use the following command to integrate it with your existing Openclaw Skills:
openclaw install error-guard
After installation, you can verify the status of your environment by running the /status command in your agent console.
error-guard Data Schema & Taxonomy
error-guard is designed with a minimal state footprint to ensure speed and security. It avoids storing sensitive payloads, prompts, or model outputs, focusing purely on lifecycle metadata.
| Field | Type | Description |
|---|---|---|
| taskId | String | Unique identifier for the current execution unit. |
| type | String | The category of the task being performed (e.g., exec, sub-agent). |
| state | Enum | Current status: active, stalled, or overdue. |
| heartbeat | Timestamp | The last recorded activity from the worker process. |
error-guard Advanced Features
- Constant-time execution for recovery commands to ensure responsiveness even during high CPU load.
- Non-LLM reasoning path ensures recovery commands work even when the model API is unresponsive.
- Support for future TTL-based task watchdogs and silence detection to automate recovery.
- Deterministic task registry reset which clears in-memory queues without data loss in external volumes.
SKILL.md
Loading
Related Openclaw Skills
Filesystem Operations
A comprehensive toolkit for advanced filesystem management, enabling recursive searches, batch operations, and deep directory analysis.
Deep Thinking Protocol
A comprehensive reasoning framework designed to guide AI agents through complex, multi-step problem-solving and architectural decision-making.
Internet Lookup Verifier
A validation tool that performs real-time internet searches to confirm facts and provide source-backed answers for AI agents.
Compound Engineering
A framework for AI agents to automatically review sessions, extract learnings, and update long-term memory to improve performance daily.
Theagora
Theagora is a marketplace and protocol for agent-to-agent service commerce featuring atomic escrow, cryptographic verification, and reputation tracking.
my-mind Manager
A structured AI management skill for capturing inspiration and streamlining the article creation lifecycle.
Featured*
