Vagrant Skill

Disposable VMs for safe testing — full sudo, Docker, Go, nested KVM, destroy and recreate. Use when you need a sandbox to build, test, or break things without affecting the host. Use when the user says "spin up a VM", "test in isolation", "run with sudo", "try something dangerous", "set up a dev environment", "I need Docker", "test firewall rules", "clean environment", or needs root access, network testing, or safe experimentation.

User Input

text

$ARGUMENTS

You MUST consider the user input before proceeding (if not empty).

Execution Instructions

This skill provides a disposable VM sandbox for safe development and testing. All privileged operations (sudo, iptables, Docker, systemd, KVM) happen inside the VM — the host is never modified.

You (the agent) should NEVER need host-level sudo. Instead, use this Vagrant workflow to get a fully provisioned, disposable VM where you have unrestricted root access.

Prerequisites

Before using this workflow, verify a provider is available:

bash

# Check Vagrant is installed
command -v vagrant

# Check for a provider (any one is sufficient)
command -v prlctl        # Parallels (Mac Apple Silicon — recommended)
vagrant plugin list | grep libvirt  # libvirt (Linux — nested KVM)
command -v VBoxManage    # VirtualBox (fallback)

See references/platform-setup.md for detailed provider installation.

---

Core Workflow

Step 1: Create a Vagrantfile in the User's Project

If the user's project does not already have a Vagrantfile, you MUST create a real, working one in their project directory. This is not a template or example — it must work end-to-end with vagrant up.

Requirements for the Vagrantfile you create:

• Base box: bento/ubuntu-24.04
• Sync the project root into the VM: config.vm.synced_folder ".", "/project", type: "rsync"
• Include all three providers (Parallels, libvirt, VirtualBox) so it works on any platform
• Provision with the actual tools the project needs (inspect the project first — look for go.mod, package.json, requirements.txt, Makefile, Dockerfile, etc.)
• Use set -euo pipefail in provisioning scripts
• Make provisioning idempotent

Here is the base Vagrantfile — you MUST customize the provisioning section based on what the project actually uses:

ruby

# -*- mode: ruby -*-
# Vagrantfile — disposable dev/test VM

VM_CPUS   = Integer(ENV["VM_CPUS"]   || 4)
VM_MEMORY = Integer(ENV["VM_MEMORY"] || 4096)

Vagrant.configure("2") do |config|
  config.vm.box = "bento/ubuntu-24.04"
  config.vm.box_check_update = false
  config.vm.hostname = "dev"
  config.vm.boot_timeout = 300
  config.ssh.forward_agent = true

  # ─── Sync project into VM at /project ─────────────────────────────────────
  config.vm.synced_folder ".", "/project", type: "rsync",
    rsync__exclude: [
      ".git/", "node_modules/", "vendor/", ".vagrant/",
      "bin/", "dist/", "build/", ".next/",
    ]

  # ─── Provider: Parallels (Mac Apple Silicon — recommended) ────────────────
  config.vm.provider "parallels" do |prl|
    prl.cpus   = VM_CPUS
    prl.memory = VM_MEMORY
    prl.update_guest_tools = true
  end

  # ─── Provider: libvirt (Linux — preferred, nested KVM) ────────────────────
  config.vm.provider "libvirt" do |lv|
    lv.cpus   = VM_CPUS
    lv.memory = VM_MEMORY
    lv.cpu_mode = "host-passthrough"
    lv.nested = true
  end

  # ─── Provider: VirtualBox (cross-platform fallback) ───────────────────────
  config.vm.provider "virtualbox" do |vb|
    vb.cpus   = VM_CPUS
    vb.memory = VM_MEMORY
    vb.customize ["modifyvm", :id, "--nested-hw-virt", "on"]
  end

  # ─── Provision: install project dependencies ──────────────────────────────
  # CUSTOMIZE THIS for the project. Inspect go.mod, package.json,
  # requirements.txt, Dockerfile, Makefile, etc. and install what's needed.
  config.vm.provision "shell", privileged: true, inline: <<-SHELL
    set -euo pipefail
    export DEBIAN_FRONTEND=noninteractive
    apt-get update -qq
    apt-get install -y -qq build-essential curl git jq ca-certificates gnupg

    # ── Docker ────────────────────────────────────────────────────────────
    if ! command -v docker &>/dev/null; then
      install -m 0755 -d /etc/apt/keyrings
      curl -fsSL https://download.docker.com/linux/ubuntu/gpg \
        -o /etc/apt/keyrings/docker.asc
      chmod a+r /etc/apt/keyrings/docker.asc
      echo "deb [arch=$(dpkg --print-architecture) \
        signed-by=/etc/apt/keyrings/docker.asc] \
        https://download.docker.com/linux/ubuntu \
        $(. /etc/os-release && echo "$VERSION_CODENAME") stable" \
        | tee /etc/apt/sources.list.d/docker.list > /dev/null
      apt-get update -qq
      apt-get install -y -qq docker-ce docker-ce-cli containerd.io
    fi
    systemctl enable --now docker
    usermod -aG docker vagrant

    # ── Add project-specific tooling below ────────────────────────────────
    # Examples (uncomment / add what the project needs):
    #
    # Go:
    #   GO_VERSION="1.24.3"
    #   curl -fsSL "https://go.dev/dl/go${GO_VERSION}.linux-$(dpkg --print-architecture).tar.gz" \
    #     | tar -C /usr/local -xz
    #   echo 'export PATH="/usr/local/go/bin:$HOME/go/bin:$PATH"' > /etc/profile.d/go.sh
    #
    # Node.js:
    #   curl -fsSL https://deb.nodesource.com/setup_22.x | bash -
    #   apt-get install -y -qq nodejs
    #
    # Python:
    #   apt-get install -y -qq python3 python3-pip python3-venv

    echo "VM ready — project synced at /project"
  SHELL
end

You MUST customize the provisioning section. Read the project to determine what it needs:

• go.mod → install Go
• package.json → install Node.js
• requirements.txt / pyproject.toml → install Python
• Dockerfile → Docker is already included above
• Network testing → add iptables dnsmasq dnsutils iproute2 net-tools
• KVM/microVM testing → add qemu-kvm libvirt-daemon-system

Uncomment the relevant blocks and add any other tooling. Do NOT leave placeholder comments in the final Vagrantfile — produce a clean, working file.

Then add .vagrant/ to the user's .gitignore if not already present:

bash

grep -qxF '.vagrant/' .gitignore 2>/dev/null || echo '.vagrant/' >> .gitignore

The Vagrantfile itself should be committed — it's reusable project infrastructure.

If the user already has a Vagrantfile, use it as-is unless they ask to modify it.

Step 2: Start the VM

bash

vagrant up

This boots the VM with the user's project synced at /project inside the VM.

Step 3: Run Commands Inside the VM

All commands use vagrant ssh -c from the host. No interactive SSH needed.

bash

# Run any command with sudo
vagrant ssh -c "sudo apt-get install -y some-package"

# Build the project
vagrant ssh -c "cd /project && make build"
vagrant ssh -c "cd /project && go test ./..."

# Docker operations
vagrant ssh -c "docker build -t myimage ."
vagrant ssh -c "docker run --rm myimage"

# Network/firewall testing
vagrant ssh -c "sudo iptables -L -n"

Step 4: Iterate on Code Changes

When you modify source on the host:

bash

vagrant rsync                                    # sync changes to VM
vagrant ssh -c "cd /project && make build"       # rebuild
vagrant ssh -c "cd /project && make test"        # test

Step 5: Tear Down

bash

vagrant destroy -f    # destroys VM completely, clean slate

---

Testing Patterns

Pattern: Build-Test-Fix Loop

bash

vagrant rsync && vagrant ssh -c "cd /project && make build"
vagrant ssh -c "cd /project && make test"
# If tests fail, fix code on host, repeat

Pattern: Docker-in-VM

bash

vagrant ssh -c "cd /project && docker build -t test ."
vagrant ssh -c "docker run --rm test"

Pattern: Network/Firewall Testing

bash

vagrant ssh -c "sudo iptables -A FORWARD -s 172.16.0.0/24 -j DROP"
vagrant ssh -c "sudo iptables -L -n -v"

Pattern: bats End-to-End Tests

Run a bats-core test suite against a live VM as proof that the system under test works. The VM must be up before running bats.

bash

vagrant up
bats test/e2e.bats      # run tests, output is the proof
vagrant destroy -f      # tear down after

Capture output to show the user:

bash

vagrant up
bats test/e2e.bats 2>&1 | tee /tmp/bats-results.txt
vagrant destroy -f
cat /tmp/bats-results.txt

bats exits non-zero on any failure — treat that as a test run failure.

Pattern: Full Reprovision (Nuclear Option)

bash

vagrant destroy -f && vagrant up

---

Configuration

Environment variables to customize the VM (set before vagrant up):

Variable	Default	Purpose
VM_CPUS	4	Number of vCPUs
VM_MEMORY	4096	RAM in MB

Example:

bash

VM_CPUS=8 VM_MEMORY=8192 vagrant up

See references/vm-contents.md for full details on VM filesystem layout and installed software.

Safety Guarantees

• No host sudo required — all privileged operations are inside the VM
• Fully disposable — vagrant destroy -f removes everything
• Idempotent provisioning — vagrant provision is safe to re-run
• Isolated networking — VM has its own network stack
• Source is rsynced — VM gets a copy; your host repo is never modified by the VM
• No persistent state — destroying the VM removes all data
• Vagrantfile is committed — reusable across sessions; .vagrant/ is gitignored

Examples

Example 1: Test iptables firewall rules without touching host network

User says: "I need to test some firewall rules before deploying to production"

Actions:

1. Create Vagrantfile with iptables dnsmasq dnsutils iproute2 net-tools provisioned
2. vagrant up
3. vagrant ssh -c "sudo iptables -A INPUT -p tcp --dport 443 -j ACCEPT"
4. vagrant ssh -c "sudo iptables -A INPUT -p tcp --dport 0:442 -j DROP"
5. vagrant ssh -c "sudo iptables -L -n -v" — verify rules look right
6. vagrant ssh -c "sudo iptables-save > /project/firewall.rules" — export if good
7. vagrant rsync to get rules file back, or vagrant destroy -f to scrap

Result: Firewall rules iterated safely. Host network never touched. Rules exportable.

Example 2: Test systemd service configuration

User says: "I need to test this systemd unit file before deploying"

Actions:

1. Create Vagrantfile with the project synced
2. vagrant up
3. vagrant ssh -c "sudo cp /project/myservice.service /etc/systemd/system/"
4. vagrant ssh -c "sudo systemctl daemon-reload && sudo systemctl start myservice"
5. vagrant ssh -c "systemctl status myservice" — check it works
6. vagrant ssh -c "sudo journalctl -u myservice --no-pager" — check logs
7. vagrant destroy -f — clean slate

Result: Service tested with real systemd, real journald. No risk to host init system.

Example 3: Docker daemon configuration and privileged port binding

User says: "I need to test a Docker compose setup that binds port 80"

Actions:

1. Create Vagrantfile with Docker CE provisioned
2. vagrant up
3. vagrant ssh -c "cd /project && docker compose up -d"
4. vagrant ssh -c "curl -sf http://localhost" — test from inside VM
5. vagrant ssh -c "docker compose logs" — check output
6. vagrant ssh -c "docker compose down" — cleanup
7. vagrant destroy -f

Result: Full Docker compose stack running with privileged ports — impossible without sudo on the host.

Example 4: Run the built-in e2e examples

Three working examples live under examples/ in this skill's directory. All follow the same pattern — boot, test, tear down:

bash

cd examples/<name>
vagrant up [--provider=<provider>]
bats test/e2e.bats 2>&1 | tee /tmp/e2e-results.txt
cat /tmp/e2e-results.txt
vagrant destroy -f

examples/nginx-hardened/ — Linux / libvirt (16 tests)

Deploys nginx + hardened iptables (INPUT DROP, allow SSH + HTTP only). Why VM: iptables -F INPUT; iptables -P INPUT DROP on the host locks you out.

bash

cd examples/nginx-hardened
vagrant up
bats test/e2e.bats 2>&1 | tee /tmp/e2e-results.txt
vagrant destroy -f

examples/mac-docker-compose/ — Mac Apple Silicon / Parallels (14 tests)

Runs a Docker Compose stack: nginx on port 80 proxying a Python JSON API. Why VM: Docker Desktop requires a commercial license; Docker CE in a VM has none of its restrictions.

bash

cd examples/mac-docker-compose
vagrant up --provider=parallels
bats test/e2e.bats 2>&1 | tee /tmp/e2e-results.txt
vagrant destroy -f

examples/windows-systemd-service/ — Windows WSL2 / VirtualBox (20 tests)

Deploys a Python HTTP server as a real systemd unit, running as a dedicated system user. Why VM: WSL2 does not run real systemd — unit files cannot be tested without a real Linux init.

WSL2 pre-flight:

bash

export VAGRANT_WSL_ENABLE_WINDOWS_ACCESS="1"
export PATH="$PATH:/mnt/c/Program Files/Oracle/VirtualBox"

bash

cd examples/windows-systemd-service
vagrant up --provider=virtualbox
bats test/e2e.bats 2>&1 | tee /tmp/e2e-results.txt
vagrant destroy -f

Troubleshooting

VM Won't Boot

Error: vagrant up hangs or times out Cause: Provider not installed or configured correctly Solution:

1. Check provider is installed: vagrant plugin list
2. Debug boot: vagrant up --debug 2>&1 | tail -50
3. Try explicit provider: vagrant up --provider=virtualbox

Source Not Synced

Error: /project directory is empty or missing inside VM Cause: rsync failed or synced_folder misconfigured Solution:

1. Re-sync: vagrant rsync
2. Check Vagrantfile has synced_folder ".", "/project", type: "rsync"

Provider Mismatch

Error: vagrant up uses wrong provider Cause: Multiple providers installed, Vagrant auto-selects Solution:

1. Check what's running: vagrant status
2. Force provider: vagrant up --provider=parallels

KVM Not Available Inside VM

Error: /dev/kvm missing inside the VM Cause: Host doesn't support nested virtualization or provider not configured Solution:

1. Ensure host has KVM: test -e /dev/kvm on host
2. Use libvirt provider with cpu_mode = "host-passthrough"
3. Mac: nested KVM is not available — use a Linux host for KVM workloads