A local development container and multi-agent swarm manager built specifically for Apple Silicon Macs.
The fastest way to install a skill directly from the registry.
npx clawhub@latest install ciderbox
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
Copy this prompt to OpenClaw to install it automatically.
Help me install ciderbox using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Get the raw skill files in a ZIP archive.
ciderbox is a powerful command-line tool designed to build and manage Apple-native container development environments. By wrapping Apple's native container CLI, it enables high-performance local container execution without relying on remote cloud runners or heavy virtualization alternatives. It is specifically engineered for local Apple Silicon Macs, providing a streamlined experience for running isolated tasks and testing environments.
Beyond basic container management, it serves as the foundational layer for managing an Orchard agent swarm, which is a local multi-agent system. This allows developers to provision Apple Container VMs, automatically integrate Openclaw Skills into each instance, and distribute complex tasks simultaneously across multiple localized agent trees.
Ensure your development machine is an Apple Silicon Mac running macOS 26 or higher, and that the Apple native container CLI is running via container system start.
# Scaffold the local project configuration file (.ciderbox.yaml)
ciderbox init
# Alternatively, auto-detect project settings and commands during initialization
ciderbox init --detect
# Run the environment doctor to ensure paths and paths for Openclaw Skills are correct
ciderbox doctor
# Execute cross-distro compilation and tests
ciderbox compile-test
# Execute a single test command inside a fresh environment
ciderbox run -- uname -a
# Tear down and destroy all active ciderbox container VMs
ciderbox chop
ciderbox relies on two primary configuration files to manage environments, data workflows, and Openclaw Skills integration:
.ciderbox.yaml)This file controls how cross-distro testing and builds are managed locally.
| Section | Property | Type | Description |
|---|---|---|---|
| compileTest | distros |
Array | A list of target Linux distributions containing name and image keys. |
parallel |
Boolean | Toggles whether multi-distro compilations run simultaneously. | |
dependencies |
Array | Required package binaries (e.g., rsync) installed before testing. |
|
command |
String | The exact test or compilation command to execute inside the container. |
.orchard.yaml)This file configures the local multi-agent environment and defines how Openclaw Skills operate inside the VMs.
name: Identifier string for the local agent swarm.trees: Total number of VM instances to provision simultaneously.image: Base Linux image (e.g., debian:bookworm).cpus: Number of processor cores allocated to each individual container VM.memory: The amount of RAM allocated per container instance.identity: The active persona designation for the agent.skills: An array tracking active Openclaw Skills injected into the agent environment.model: The AI LLM target path (e.g., ollama/llama3.1).command: The specific runtime string utilized to initialize the background agent loop.secrets.envFile: The path pointing to local credential storage (defaults to .orchid.env).workspace.sync: A boolean flag determining if your local project codebase mirrors inside the running VM context.--detect flag that parses your directory structure to automatically fill in specific project-matching dependencies and test commands.ciderbox orchard secrets push without leaking them to the host logging framework.Loading
An AI-powered decision-making toolkit that translates historical monetary policy traps, central banking crises, and macroeconomic dogma into actionable insights for modern decision-makers.

An agent-to-agent direct messaging skill that enables AI agents to securely register handles, send and receive DMs, and manage conversations on The Colony infrastructure.

An automated, headless connection tool utilizing Google service accounts to seamlessly read and write Google Sheets, Docs, Drive, and Calendars without manual browser OAuth flows.

A local market intelligence skill that analyzes Google Maps CSV exports to identify competitor vulnerabilities, underserved niches, and digital marketing gaps.

Luban is an artisan-style quality-assurance skill designed to audit, refine, and optimize raw agent workflows into verifiable, production-ready assets.

MeshKore is an open-mesh router that enables AI agents to discover, compare, and interact with over 69,000 specialized service agents worldwide.








































