A secure clean room environment for executing untrusted scripts and commands within isolated Docker containers.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-securitysandbox
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 agent-securitysandbox 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 Multi-Agent Security Sandbox (MASS)?
The Multi-Agent Security Sandbox (MASS) is a critical security layer designed to protect host systems from potentially malicious or unstable code. By leveraging Docker with hardened configurations, it creates a restricted execution environment where AI agents can safely test third-party scripts or unknown commands without risking the integrity of the primary workstation.
Integrating this tool into your Openclaw Skills library ensures that every execution is contained within a non-root environment with zero network access. It is an essential utility for developers and AI researchers who frequently interact with external, unverified codebases or complex automated workflows.
Multi-Agent Security Sandbox (MASS) Use Cases
- Safely executing scripts from untrusted or third-party sources.
- Testing and debugging new agent skills that require system-level access.
- Verifying the behavior of suspicious code in a controlled, isolated environment.
- Preventing accidental host system modifications during automated CLI tasks.
How Multi-Agent Security Sandbox (MASS) Works
- Input Identification: The agent or user identifies a command, script, or binary that requires isolated execution.
- Environment Initialization: The MASS script triggers a transient Docker container based on a hardened security profile.
- Restricted Execution: The command is executed within the container under a non-root user with all Linux capabilities dropped and networking disabled.
- Log Collection: Stdout and stderr are captured and returned to the host for review.
- Cleanup: The container is immediately removed, ensuring no persistence or side effects remain on the system.
Multi-Agent Security Sandbox (MASS) Setup
Ensure you have Docker installed on your host system. To use this feature within your Openclaw Skills collection, navigate to the skill directory and use the provided wrapper script.
# Check for docker dependency
docker --version
# Execute a command in the sandbox
./scripts/mass "python3 suspicious_script.py"
Multi-Agent Security Sandbox (MASS) Data Schema & Taxonomy
The MASS skill manages isolation through specific configuration files and ephemeral container states:
| Component | Description |
|---|---|
scripts/mass |
The primary execution wrapper for sandboxed commands. |
seccomp.json |
Security profile defining allowed and blocked system calls. |
Dockerfile |
Defines the environment, installed tools, and the sandboxuser. |
| Container State | Stateless; all modifications are wiped post-execution via --rm. |
Multi-Agent Security Sandbox (MASS) Advanced Features
- Cap-Drop Isolation: Drops all default Linux capabilities (
--cap-drop=ALL) to prevent privilege escalation. - Network Air-Gapping: Disables all inbound and outbound network traffic by default.
- Seccomp Filtering: Custom syscall filtering to block sensitive kernel interactions.
- Ephemeral Runtime: Automatic container destruction ensures a fresh environment for every single execution call.
SKILL.md
Loading
Related Openclaw Skills
Bidding Document Parser
An intelligent assistant designed to parse complex bidding documents and generate structured, formatted analysis reports with original page citations.
Social Video Distill
An AI-first delegation tool designed to distill short-form social media video into transcripts, summaries, and creative insights using the fastest available cloud or browser routes.
AI Cardiac Rehab Management System
A clinical-grade cardiac rehabilitation web application designed for automated risk stratification and personalized recovery planning.
Medical Bill Review Kit
A specialized tool for organizing medical billing documents, identifying common errors, and preparing structured scripts for insurance and provider negotiations.
Flight Disruption Compensation Kit
A comprehensive self-help toolkit for air passengers to document disruptions and prepare structured compensation claims under global aviation regulations.
Sichuan Cuisine Skill
A comprehensive digital knowledge base for mastering the art of Sichuan cuisine, covering everything from Pixian doubanjiang techniques to 45+ iconic recipes.
Featured*
