panic

To halt AI operations during critical system failures, you can trigger an emergency stop via a system-governor agent, which initiates a controlled pause to prevent cascading damage and unsafe behavior without losing data.

An emergency stop protocol for runtime governance contexts defines a controlled halt mechanism using a system-governor to pause AI operations safely. It establishes a lockfile to signal the halt and requires deliberate user action to resume operations.

To resume system operations after triggering an emergency halt, you must perform an explicit resume procedure. This deliberate user action removes the HALT lockfile and restores operations, reducing the risk of accidental continuation.

The emergency halt mechanism works within runtime governance contexts that support a system-governor agent. It relies on creating a lockfile at the .claude/HALT directory to signal the halt and manage the controlled shutdown safely.

The best way to prevent unsafe behavior during critical failures is to use a system-governor to issue a controlled shutdown. This creates an audit trail via a lockfile and requires explicit user action to resume, ensuring safety.

You should not use an automated emergency stop if your environment lacks a system-governor runtime context, as the halt mechanism depends on creating a lockfile at .claude/HALT and requires explicit user action to safely resume operations.