How do I monitor the progress of a delegated task?

You can monitor tasks by using the 'bdy sandbox exec status' and 'bdy sandbox exec logs' commands, which allow you to poll for live output or wait for completion.

Does the sub-agent have access to my local files?

No, the sub-agent is isolated in the sandbox. You must explicitly copy files to the sandbox or use a snapshot to provide the sub-agent with the necessary context.

Can I run multiple Claude agents simultaneously?

Yes, you can create multiple sandboxes and delegate different tasks (or the same task for comparison) to each one, enabling high-concurrency AI workflows.

What is YOLO mode in Sandbox Agent?

YOLO mode allows sub-agents to execute any command without asking for user permission prompts. This is safe because the agent is contained within an isolated Buddy Sandbox, protecting your host machine.

What resource profiles are available for sandboxes?

You can choose between 2x4 (light), 4x8 (standard), and 8x16 (heavy) CPU and RAM configurations to match the requirements of your task.

Sandbox Agent Task Delegation

Name: Sandbox Agent Task Delegation
Author: sztwiorok

bysztwiorok

0•

개발자 도구

Delegates AI tasks to autonomous Claude agents running in isolated Buddy Sandbox environments for safe, parallel execution.

The Sandbox Agent skill enables Claude to spawn sub-agents in secure, isolated environments to perform complex coding and system tasks without risking the primary workspace. It is specifically designed for scenarios requiring 'YOLO mode' execution where permissions are pre-cleared, parallel processing of multiple tasks, or side-by-side comparison of different AI solutions. By decoupling the execution environment from the main workspace, developers can safely run experimental scripts, automate multi-step workflows, and scale their productivity through multi-agent orchestration.

주요 기능

010 GitHub stars

02Live Progress Monitoring: Poll logs and status updates from sub-agents in real-time.

03YOLO Mode: Execute commands without permission prompts in isolated sandboxes.

04Parallel Agent Processing: Spin up multiple sandboxes to run sub-tasks simultaneously.

05Multi-Agent Comparison: Run the same prompt through different agents to compare and merge results.

06Isolated Task Execution: Protect your local environment from experimental code or destructive commands.

사용 사례

01Running untrusted scripts or experimental refactors in a clean, throwaway environment.

02Delegating long-running documentation or linting tasks to sub-agents while continuing active development.

03Simultaneously generating and running unit tests across multiple OS versions or snapshots.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add sztwiorok/buddy-plugin sandbox-agent

For use in Claude.ai and ChatGPT

Download Skill