write-subagent-todos

To coordinate parallel execution, decompose your complex objective into structured subtasks and assign each to specialized subagents. The main process plans and coordinates the workflow while the subagents perform the hands-on work, ensuring safe coordination and synthesizing the final results.

The best way to decompose a project is to create a multi-step todo list where each step is assigned to the most suitable subagent. You need at least 3 distinct steps with clear agent responsibilities, allowing the main process to plan and assign subagent types dynamically.

Yes, you can use custom agents for specialized tasks. User-defined agents take priority over built-in ones, and the system automatically discovers and dynamically loads custom subagents from your project configurations to execute the assigned work.

Yes, subagent task management works for non-coding cross-domain projects. It flexibly supports research, documentation, testing, and other multi-domain tasks that require parallel execution and workload distribution across specialized agents.

You should not use subagent coordination for tasks that require fewer than 3 distinct steps, lack clear agent responsibilities, or do not benefit from parallel execution. It is designed specifically for complex objectives requiring structured decomposition and safe coordination.