Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Control-plane skill for OpenClaw that prevents agent deadlocks, freezes, and unrecoverable states. Provides three non-LLM commands (/status, /flush, /recover) to inspect task state, cancel active work, and reset the system without restarting the container. Intended for long-running workloads, sub-agents, benchmarks, and background monitors.
Restores a hung agent to a clean state without a container restart, avoiding cold-start overhead and state loss that a full restart requires.
When to use it
- Recovering a frozen agent without restarting the container
- Canceling runaway background monitors mid-execution
- Inspecting stalled sub-agent task state during a benchmark run
- Flushing message queues when the system stops responding
- Resetting control-plane state after a deadlock
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: the agent is frozen and not responding to new tasks
- 1Run /status to check active tasks, last heartbeats, and stalled flags
- 2Identify which task is overdue or silent
- 3Run /flush to cancel all active tasks, kill exec sessions, and clear the message queue
- 4Run /recover to reset control-plane state and optionally reload skills without a container restart
System returns to a clean idle state with no container restart required