prose

Multi-agent AI orchestration with persistent state is handled by defining agents, sessions, and state management within a single frontmatter-driven skill unit. This centralizes coordination and enables repeatable, auditable executions across long-running tasks.

Parallel execution blocks for multiple AI agents are supported within a single session. The skill unit defines parallel tasks and coordinates agent sessions, allowing concurrent operations while maintaining centralized state management.

The captain-pattern for AI agent orchestration is a multi-agent coordination strategy where a lead agent directs workflow execution. You need it for complex, end-to-end workflows requiring automated coordination, checkpointing, and parallel task delegation.

To start automating complex AI workflows, provide a minimal program that defines two agents and a small parallel task. This quick start approach lets you observe coordinated sessions and validate the orchestration logic before scaling.

This AI orchestration skill integrates with optional scripts, references, and assets directories. These extensible runtime components allow you to execute deterministic tasks, attach documentation, and manage artifacts alongside the workflow.

VM skill packs for long-running AI tasks require checkpointing to maintain state persistence. While they enable auditable executions, complex multi-agent coordination depends on the VM runtime environment and proper frontmatter configuration to avoid state loss.