Shadowbox — AI skill for Claude Code
Zero-trust copy-on-write sandbox and interactive time-machine for AI coding agents.
How to install Shadowbox
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open Nari-Ab/shadowbox and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Shadowbox does
Zero-trust copy-on-write sandbox and interactive time-machine for AI coding agents.
Alternatives in AI
- Coding Agent Template — Multi-agent AI coding platform powered by Vercel Sandbox and AI Gateway 1.8k ★
- Vibe Coding Toolkit — A curated, battle-tested AI-coding toolkit: Claude Code plugins, subagent orchestration, quality gates, and re 297 ★
- Learn Skills.Dev — Curated high-quality AI Agent Skills 203 ★
README
Zero-trust copy-on-write sandbox and time-machine for AI coding agents
[](https://github.com/Nari-Ab/shadowbox/actions/workflows/ci.yml) [](https://crates.io/crates/shadowbox) [](https://www.npmjs.com/package/shadowbox-cli) [](https://github.com/Nari-Ab/shadowbox/releases) [](LICENSE)
ShadowBox runs autonomous AI coding agents (Claude Code, Cursor, Aider, OpenHands) inside an ephemeral copy-on-write scratchpad on macOS and Linux. It strips secret credentials before the agent process spawns, isolates file mutations on an APFS zero-copy layer, and lets you review or discard changes before anything touches your real repository.
# curl (macOS / Linux):
curl -fsSL https://raw.githubusercontent.com/Nari-Ab/shadowbox/main/install.sh | bash
# npx (instant, zero install):
npx shadowbox-cli claude
# cargo:
cargo install shadowbox
# run:
shadowbox claude --dangerously-skip-permissions
Needs macOS (APFS) or Linux.
What a session looks like:
shadowbox session — zero-trust copy-on-write runtime
2 file(s) modified in sandbox:
● [A] src/auth/jwt.ts (+38 lines)
● [M] package.json (+1 line)
SECRETS BLOCKED (3)
● AWS_SECRET_ACCESS_KEY — stripped from process env
● GITHUB_TOKEN — stripped from process env
● .env — masked with [SHADOWBOX_REDACTED]
HOST REPOSITORY
● 100% untouched, 0 unapproved writes
CONTROLS
[c] Commit
Related Skills
Kaut
KAUT — Knowledge Actualization Under Trust: self-maintaining, AI-first project knowledge for AI agents. Source
Cowback
🐄 The missing Ctrl+Z for AI Agents. Zero-cost undo powered by Copy-on-Write.
Tunnelcode
Run an AI coding agent on your own machine and drive it from a browser. Pair your phone with a QR code, watch
Wiggle
A working copy of the sandbox that Claude (claude.ai) runs code in.
Windbg Ext MCP
WinDbg-ext-MCP bridges your favorite LLM client (like Cursor, Claude, or VS Code) with WinDbg, enabling real-t
Frankenterm
Terminal hypervisor for AI agent swarms: real-time pane capture, state-machine pattern detection, and a JSON A
Related Agents
Interactive Troubleshooter
Automated diagnostic agent that runs checks, identifies issues, and provides targeted fixes for Docker sandbox
Sf Security Engineer
Security Engineer of the Software Factory cell. Use to run SAST/dependency/secret scans, verify zero-trust sec
Security Boss
You need to identify vulnerabilities, audit auth flows, review code for security flaws, or threat-model a syst