execution-engine

Automate multi-step task execution by running an autonomous loop that manages context and approvals. This skill executes decomposed tasks sequentially, maintaining state and retry logic to prevent agent drift and premature closure.

State management for interrupted workflow execution works by maintaining execution state and progress logs. This allows graceful interruption and resumption of multi-step implementation workflows without losing completed task data or retry budgets.

Yes, you need a stop hook to manage task iteration budgets. The stop hook intercepts session exits, manages context, and enforces human approval gates to ensure reliable unattended task completion during autonomous execution.

Yes, you can pause autonomous code execution for human approval gates. The execution loop runs tasks sequentially with fresh context, respecting defined review gates to pause for human approval before proceeding with implementation workflows.

You should not use an autonomous execution loop for single-step tasks or workflows lacking defined review gates. It is specifically designed for multi-step implementation requiring state persistence, retry logic, and stop hook integration to manage iteration budgets.