OpenProse is a specialized programming language that transforms LLMs into high-fidelity virtual machines capable of orchestrating complex multi-agent sessions.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install open-prose
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install open-prose using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is OpenProse AI Virtual Machine?
OpenProse is built on the principle that LLMs are powerful simulators. When provided with a detailed system description, an AI session doesn't just describe a workflow—it simulates the execution of that workflow. By using this skill within the ecosystem of Openclaw Skills, developers can treat the AI as a virtual machine (VM) that runs structured .prose programs. This enables a robust environment where simulation with sufficient fidelity becomes implementation.
This skill provides the runtime mapping necessary to bridge the gap between abstract AI instructions and concrete actions like file I/O, remote data fetching, and sub-agent spawning. Whether you are building a custom browser from scratch or managing a parallel code review pipeline, OpenProse offers the framework to script AI behaviors with precision and repeatability.
OpenProse AI Virtual Machine Use Cases
- Orchestrating complex multi-agent workflows from a single script to handle multi-step engineering tasks.
- Creating reusable and shareable AI automation programs that can be executed across different environments.
- Simulating specialized tools and environments, such as code compilers or web browsers, within an AI session.
- Managing long-running AI tasks with persistent state that can be resumed or audited.
- Automating production-grade pipelines for content generation, bug hunting, and feature development.
How OpenProse AI Virtual Machine Works
- The user invokes a prose command (e.g., prose run or prose compile), which activates the skill via the Openclaw Skills runtime.
- The agent loads the core prose.md specification to transform its current session into the OpenProse Virtual Machine.
- The system parses the target .prose file and resolves any dependencies, including local examples or remote programs from the registry.
- The VM executes the program logic, utilizing the sessions_spawn tool to manage sub-agents and read/write tools for state persistence.
- Real-time narration tracks the execution state, including bindings and logical positions, ensuring the workflow remains transparent and controllable.
- Upon completion, the state is saved to the configured backend (filesystem, SQLite, or PostgreSQL) for future reference or resumption.
OpenProse AI Virtual Machine Setup
To begin using the OpenProse environment within Openclaw Skills, ensure your workspace is properly initialized.
# Initialize your project and check the current version
prose help
# Run a migration if you are moving from a legacy setup
prose update
Configure your environment by creating a .prose/.env file in your project root to store necessary variables:
# Example .env configuration
OPENPROSE_TELEMETRY=enabled
OPENPROSE_POSTGRES_URL=postgresql://user:pass@localhost:5432/prose
You can immediately test the installation by running one of the bundled examples:
prose run examples/01-hello-world.prose
OpenProse AI Virtual Machine Data Schema & Taxonomy
The OpenProse skill manages data through a structured hierarchy in the user's workspace and home directory to ensure portability and persistence.
| Location | Data Type | Purpose |
|---|---|---|
.prose/runs/ |
Runtime State | Stores session-specific logs, snapshots, and execution history. |
.prose/agents/ |
Project Agents | Local repository for persistent agents scoped to the current project. |
~/.prose/agents/ |
Global Agents | Cross-project storage for persistent agents available to all sessions. |
.prose/.env |
Configuration | Key-value store for environment variables and feature toggles. |
*.prose |
Source Code | User-defined programs and orchestration scripts. |
OpenProse AI Virtual Machine Advanced Features
- Multi-Agent Orchestration: Leverage the sessions_spawn protocol to coordinate multiple specialized agents within a single program logic.
- Remote Registry Support: Run programs directly from URLs or the official registry using shorthand syntax like prose run handle/slug.
- Pluggable State Engines: Switch between filesystem, in-context, SQLite, or PostgreSQL backends to match your scale and concurrency needs.
- Integrated Validation: Use the built-in compiler.md logic to validate program syntax and logic before initiating expensive AI sessions.
- Intelligent Narration: Automated tracking of the VM's internal state using the [Position], [Binding], and [Success] narration markers for easier debugging.
SKILL.md
Loading
Related Openclaw Skills
Tunneling (TinyFish)
A seamless SSH-based tunneling skill to create public HTTPS URLs for local applications without any signup or software installation.
Manifold Markets Prediction Trading
A comprehensive interface for searching, analyzing, and executing trades on Manifold Markets prediction platforms.
Kagi Search
A lightweight CLI tool providing high-quality, privacy-focused web search results for AI agents and developers.
activity-notifier
A real-time status broadcasting tool that keeps users informed of an AI agent's progress via Discord.
Exa Search & Research MCP
A powerful neural search integration that enables AI agents to perform advanced web crawling and deep research using the Exa protocol.
Image Generator Skill
A local image generation workflow supporting Stable Diffusion 1.5 and SDXL models integrated with non-blocking subagent execution.
Featured*
