daaain/devcontainer-security

Guide for setting up secured VS Code dev containers for coding agents. Use when creating or hardening a DevContainer to sandbox Claude Code or other coding agents, configuring Docker socket proxies, handling VS Code IPC escape vectors, setting up git worktree support, or verifying security controls. Covers threat model, three-layer defence architecture, Node.js/pnpm setup, and verification testing.

Quality

95%

Does it follow best practices?

Impact

Pending

No eval scenarios have been run

Securityby

Advisory

Suggest reviewing before use

name:: devcontainer-security
description:: Guide for setting up secured VS Code dev containers for coding agents. Use when creating or hardening a DevContainer to sandbox Claude Code or other coding agents, configuring Docker socket proxies, handling VS Code IPC escape vectors, setting up git worktree support, or verifying security controls. Covers threat model, three-layer defence architecture, Node.js/pnpm setup, and verification testing.
metadata:: {"version":"1.1.1"}

Secured VS Code Dev Containers for Coding Agents

Name: daaain/devcontainer-security
Rating: 95 (1 reviews)
Author: daaain

Set up a hardened VS Code DevContainer that sandboxes coding agents while maintaining full development capability.

When to Use This Skill

Setting up a new DevContainer for coding agent use
Hardening an existing DevContainer against escape vectors
Adding git worktree support for parallel development
Setting up sibling Docker services (databases, emulators)
Verifying security controls are working
Setting up Node.js / pnpm in a DevContainer

Threat Model

We're protecting against three things:

Supply chain attacks - malicious npm packages executing code during install or at runtime
Prompt injection - malicious content convincing the agent to run harmful commands
Agent mistakes - unintentional destructive actions

The goal is to limit blast radius, not eliminate all risk.

Architecture Overview

┌─────────────────────────────────────────────────┐
│                  Host Machine                   │
│  ┌───────────────────────────────────────────┐  │
│  │              Docker Engine                │  │
│  │  ┌──────────────┐  ┌──────────────────┐   │  │
│  │  │ docker-proxy │◄─│  claude-code     │   │  │
│  │  │ (read-only)  │  │  (DevContainer)  │   │  │
│  │  └──────┬───────┘  └──────────────────┘   │  │
│  │         ▼                                 │  │
│  │  ┌──────────────┐                         │  │
│  │  │ Docker Socket│                         │  │
│  │  └──────────────┘                         │  │
│  └───────────────────────────────────────────┘  │
└─────────────────────────────────────────────────┘

Security Controls Summary

Control	What It Blocks	How
Docker socket proxy	Container escape	Read-only API proxy (POST=0, EXEC=0)
No sudo	Privilege escalation	Not installed in image
Drop all capabilities	Kernel attack surface	`cap_drop: [ALL]` in docker-compose.yml
No new privileges	Setuid/setgid escalation	`security_opt: no-new-privileges:true`
No SSH keys	Git push / code exfil	Keys not mounted, agent socket deleted
VS Code IPC hardening	Host command execution	Three-layer env var + socket cleanup
No credential injection	Docker/git credential leaks	VS Code settings disabled

Documentation Index

Document	Contents
SECURITY-HARDENING.md	Three-layer defence against VS Code escape vectors, the `.bashrc` non-interactive shell subtlety
DOCKER-PROXY.md	Docker socket proxy setup, sibling container communication
NODE-SETUP.md	Node.js + pnpm Dockerfile patterns, startup commands
WORKTREE-SUPPORT.md	Git worktree support, dynamic container naming, isolated volumes
VERIFICATION.md	How to verify security controls, integration testing with devcontainers-cli, automated test scripts

Quick Start — Minimal Secured DevContainer

Three files are needed. See each sub-document for detailed explanations.

.devcontainer/devcontainer.json:

{
  "name": "Secured Dev Container",
  "dockerComposeFile": "docker-compose.yml",
  "service": "app",
  "workspaceFolder": "/app",
  "remoteUser": "vscode",
  "shutdownAction": "stopCompose",
  "remoteEnv": {
    "SSH_AUTH_SOCK": "",
    "GPG_AGENT_INFO": "",
    "BROWSER": "",
    "VSCODE_IPC_HOOK_CLI": null,
    "VSCODE_GIT_IPC_HANDLE": null,
    "GIT_ASKPASS": null,
    "VSCODE_GIT_ASKPASS_MAIN": null,
    "VSCODE_GIT_ASKPASS_NODE": null,
    "VSCODE_GIT_ASKPASS_EXTRA_ARGS": null,
    "REMOTE_CONTAINERS_IPC": null,
    "REMOTE_CONTAINERS_SOCKETS": null,
    "REMOTE_CONTAINERS_DISPLAY_SOCK": null,
    "WAYLAND_DISPLAY": null
  },
  // postStartCommand: clean up sockets created before VS Code attaches
  "postStartCommand": "find /tmp -maxdepth 2 \\( -name 'vscode-ssh-auth-*.sock' -o -name 'vscode-remote-containers-ipc-*.sock' -o -name 'vscode-remote-containers-*.js' \\) -delete 2>/dev/null || true",
  // IPC socket cleanup (vscode-ipc-*.sock, vscode-git-*.sock) is handled by a background
  // loop in the Docker Compose command — postAttachCommand is unreliable for background
  // processes due to VS Code's cgroup-based lifecycle cleanup. See SECURITY-HARDENING.md.
  "customizations": {
    "vscode": {
      "settings": {
        "dev.containers.dockerCredentialHelper": false,
        "dev.containers.copyGitConfig": false
      }
    }
  }
}

.devcontainer/docker-compose.yml: See DOCKER-PROXY.md.

.devcontainer/Dockerfile: See NODE-SETUP.md or adapt for your language.

Accepted Risks

These are trade-offs for development usability:

Risk	Why Accepted
Network egress (data exfiltration)	Development requires internet access
Workspace write access	Essential for development; git tracks changes
Agent credentials readable	Token is revocable; limited blast radius
Environment variables (.env)	Development requires env vars, no production keys

scripts

DOCKER-PROXY.md

NODE-SETUP.md

SECURITY-HARDENING.md