prose

Executes OpenProse programs that orchestrate multi-agent AI workflows through a simulated virtual machine.

1|Updated Mar 3, 2026
One-click install
npx skills add https://github.com/Cyandex/ErnOS --skill prose-cyandex
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: prose
Source: https://github.com/Cyandex/ErnOS/tree/main/extensions/open-prose/skills/prose
Command: npx skills add https://github.com/Cyandex/ErnOS --skill prose-cyandex

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Coordinating multiple AI agents for complex tasks usually requires external orchestration frameworks, custom code, and infrastructure. OpenProse treats the AI session itself as a virtual machine, letting you write structured .prose programs that spawn subagents, manage state, and handle control flow without any external dependencies. ## Core Features & Use Cases - Program Execution: Run .prose files locally, from URLs, or from the p.prose.md registry using prose run, with the AI embodying the OpenProse VM. - Multi-Agent Orchestration: Define agents with models, prompts, skills, and permissions; coordinate them with parallel blocks, loops, pipelines, and persistent memory via resume:. - Compilation & Validation: Validate .prose syntax with prose compile before execution, and migrate legacy workspaces with prose update. - Flexible State Management: Choose between filesystem (default), in-context, SQLite, or PostgreSQL state backends for run persistence. - Use Case: Write a program that fans out five research sessions in parallel, synthesizes the results, and loops until a quality condition is met—then run it with a single prose run command. ## Quick Start Run the hello world example by asking the assistant to execute prose run examples/01-hello-world.prose.

Frequently Asked Questions about prose

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run an OpenProse program?▼

Use the command prose run followed by a local .prose file path, a URL, or a registry reference like handle/slug. The AI loads the VM specification from prose.md and executes the program by spawning subagent sessions.

How do I orchestrate multiple AI agents from a script?▼

Define agents with model, prompt, skills, and permissions in a .prose file, then invoke them with session statements. Use parallel blocks for concurrent execution, context properties to pass outputs between sessions, and resume for persistent agent memory.

What is the difference between OpenProse and LangChain or CrewAI?▼

LangChain and CrewAI are external orchestration libraries that coordinate agents from outside. OpenProse runs inside the AI session itself, using the session as the container, so .prose files remain portable across AI assistants with zero external dependencies.

Which AI assistants support OpenProse programs?▼

Claude Code, OpenCode, and Amp are supported. Any harness running a sufficiently capable model with subagent primitives is considered Prose Complete and can execute .prose programs.

How do I validate a .prose file without running it?▼

Use prose compile followed by the file path. This loads compiler.md, which checks syntax against the grammar, validates that references resolve, and flags ambiguous constructs without executing any sessions.

What state backends does OpenProse support?▼

OpenProse supports filesystem state (default, stored in .prose/runs/), in-context state for simple programs, and experimental SQLite and PostgreSQL modes. SQLite requires the sqlite3 CLI; PostgreSQL requires psql and a configured OPENPROSE_POSTGRES_URL connection string.