1. Read the configuration file to determine the worker agent to use.
2. Create a plan file with a checklist of concrete actions.
3. Write the worker prompt to a temporary file.
4. Create a wrapper script to execute the worker agent with the prompt file.
5. Spawn the worker agent as a background task.
6. Report the task ID, background task ID, agent used, and a summary of the plan to the user.
7. Monitor progress by reading the plan file and reporting the status of each checklist item.
8. Handle blocked items by surfacing questions to the user and spawning new workers with the answers.
SynapticSkills