prose

To orchestrate multi-agent AI workflows, you write structured programs that simulate AI sessions and define how agents interact. This Skill interprets those programs within a virtual machine to coordinate tasks like code analysis and bug detection.

AI session simulation is the process of mimicking AI interactions through structured code. This Skill interprets your code to spawn subagents and execute tasks within a virtual machine, allowing modular composition of agent behaviors.

Yes, you need Claude Code or a similar AI framework to execute these multi-agent workflow simulations. The Skill relies on these frameworks to interpret the structured programs and manage the virtual machine operations.

Yes, you can use different AI models in the same multi-agent workflow. The Skill supports modular composition of agents, allowing you to combine various models like Opus and Haiku for distinct tasks such as sentiment analysis or code generation.

To get started with simulating AI sessions, run the '01-hello-world.prose' example included in the Skill. This demonstrates the basic structure of a program and shows how to define agent behaviors and interactions.

A limitation of using structured programming for AI agent orchestration is the dependency on specific frameworks like Claude Code to execute. Additionally, complex workflows require careful modular composition to avoid interaction conflicts between spawned subagents.