openclaw - 💡(How to fix) Fix /stop doesn't interrupt busy agents immediately — waits for current action to finish

Bug Description

When a user sends /stop, the agent should stop immediately. However, /stop does not interrupt a busy agent right away. The agent continues running until the current action (tool call, thinking, etc.) finishes, and only then processes the stop command.

This affects any agent that is mid-action — not just sub-agents.

Steps to Reproduce

1. Trigger any agent to perform a longer tool call or chain of actions
2. While still executing, send /stop

Expected Behavior

Agent stops immediately, or at minimum visibly acknowledges the stop command (e.g. "останавливаюсь") and halts further actions.

Actual Behavior

Agent continues working for 30+ seconds, finishing whatever action was in progress, and only then processes /stop. From the user perspective, /stop appears broken or ignored.

Version

OpenClaw 2026.5.12 (f066dd2)