sandbox0

Use this skill when an AI agent developer wants to add Sandbox0 sandboxing to their own agent through the CLI or SDK, or needs help choosing templates, contexts, volumes, network policy, ports, webhooks, or self-hosted deployment. It uses only files bundled inside the skill.

install

source · Clone the upstream repo

git clone https://github.com/sandbox0-ai/sandbox0

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/sandbox0-ai/sandbox0 "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/sandbox0" ~/.claude/skills/sandbox0-ai-sandbox0-sandbox0 && rm -rf "$T"

manifest: skills/sandbox0/SKILL.md

source content

Sandbox0

Use this skill for Sandbox0 product usage and integration guidance for AI agent developers.

Start here

Identify the user's integration goal.
Read
```
references/integration-map.md
```
.
Read only the relevant page under
```
references/docs-src/
```
.

Typical triggers

Add Sandbox0 to an AI agent project
Pick between REPL and one-shot execution
Persist workspace state across runs
Design a template or custom image
Restrict network access
Expose a preview server or local web app
Receive sandbox events through webhooks
Decide whether self-hosted Sandbox0 is needed

Working rules

Prefer s0 CLI- and SDK-oriented guidance over internal implementation details.
Recommend concrete compositions when helpful, for example
```
REPL + Volume
```
for persistent coding sessions.
Use architecture detail only when it changes the user's design choice, security model, or deployment shape.

Which references to read

Topic routing and common solution shapes:
```
references/integration-map.md
```

Product docs and examples:

references/docs-src/<section>/.../page.mdx

Table of contents:
```
references/docs-src/manifest.json
```