sandbox-agent

Deploy and manage a sandbox-agent server for AI coding agents.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/andreasasprou/polaris --skill sandbox-agent-andreasasprou
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: sandbox-agent
Source: https://github.com/andreasasprou/polaris/tree/main/.agents/skills/sandbox-agent
Command: npx skills add https://github.com/andreasasprou/polaris --skill sandbox-agent-andreasasprou

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Deploying and integrating AI coding agents across local and cloud sandboxes is complex and error prone; this Skill consolidates deployment, credential wiring, agent installation, session lifecycle, and streaming event handling so developers can run coding agents reliably.

Core Features & Use Cases

  • Server deployment & configuration: Guidance for running sandbox-agent server locally or in cloud sandboxes and configuring tokens and CORS.
  • Session and SDK management: Create, resume, destroy sessions and manage models, modes, and thought levels via SDK or HTTP API.
  • Event streaming & HITL: Stream turn/item deltas, handle permission requests and questions for human-in-the-loop workflows, and ingest persisted transcripts.
  • Agent lifecycle & testing: Install or lazily install agents (Claude, Codex, OpenCode, Amp), use the mock agent for offline testing, and debug with the Inspector UI.
  • Use case: Automate repository summarization, run automated PR reviews, or power an interactive chat UI that streams incremental coding suggestions.

Quick Start

Start a sandbox-agent server with your provider keys and create a session to prompt the agent to summarize the repository.

Frequently Asked Questions about sandbox-agent

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

FAQPage Schema
How do I orchestrate AI coding agents in a cloud sandbox?▼

You can orchestrate AI coding agents in a cloud sandbox by deploying a sandbox agent server to manage sessions, install agents, and stream events via SDK or HTTP API.

Does sandbox agent session management work with E2B, Daytona, and Docker?▼

Yes, sandbox agent session management supports local and cloud sandboxes including E2B, Daytona, Docker, and Vercel for creating, resuming, and destroying agent sessions.

How do I handle human-in-the-loop permission requests during event streaming?▼

Human-in-the-loop permission requests are handled during event streaming by intercepting turn and item deltas to pause execution and surface questions or permission prompts to the user.

Can I test agent lifecycle management offline without provider API keys?▼

Yes, you can test agent lifecycle management offline by using a built-in mock agent, which allows you to debug sessions and workflows without requiring real provider credentials.

What is the best way to integrate a streaming coding agent into a chat UI?▼

The best way to integrate a streaming coding agent into a chat UI is to use the SDK to stream incremental event deltas and manage the session lifecycle through the sandbox agent server.

Which AI coding agents can I install and manage through a sandbox server?▼

You can install and manage agents like Claude, Codex, OpenCode, and Amp through the sandbox server, either by pre-installing them or triggering lazy installation during session runtime.