factory-basics

Multi-agent workflow orchestration coordinates subagents by spawning them to handle parallel fan-out tasks and sequential pipelines, then synthesizing their results into a single output using the factory global and per-agent prompt routing.

Prompt routing controls agent behavior by separating the systemPrompt, which defines the agent's role, from the concrete prompt, which assigns the specific task, ensuring clearly separated responsibilities within the orchestration pipeline.

TypeScript supports building end-to-end automation pipelines by using the factory global to spawn subagents, route prompts, coordinate parallel and sequential tasks, and synthesize the final results.

The best way to structure fan-out and pipeline patterns is applying orchestration patterns where subagents execute parallel tasks, followed by sequential pipelines that pass synthesized results through defined per-agent prompts.

No external dependencies are required to run pi-factory scripts for multi-agent orchestration, as the Skill relies solely on the factory global to spawn subagents and coordinate results without additional packages.

You should not use a fan-out orchestration pattern when tasks require strict sequential execution, as fan-out patterns spawn parallel subagents that may return results out of order before pipeline synthesis.