What problem does it solve? Before sending commands, scripts, or troubleshooting steps to a customer, engineers need proof they actually work. This Skill runs action plans against real Azure lab VMs (Linux and Windows) and returns a per-step verification report, catching broken syntax, wrong targets, and destructive operations before they reach production. ## Core Features & Use Cases - Command classification and routing: Parses numbered plans or raw commands and routes each step to the right target — local az CLI, Linux VM via run-command/SSH, Windows VM via PowerShell, Azure Portal via Playwright, or ARM REST API. - Scenario labs: Builds multi-resource experiments from scratch (image → VM → restore point → disk → new VM) in a dedicated resource group with a teardown manifest, cheapest SKUs, and marker-file verification. - Safety gates: Blocks destructive commands without confirmation, refuses to touch non-lab subscriptions, and declines platform-side issues (unexpected reboots, hardware failures) that a lab cannot reproduce. - Use Case: A support engineer wants to confirm that resizing a VM to Standard_D2as_v4 preserves accelerated networking before recommending it to a customer — the Skill performs the resize on a lab VM, verifies boot and NIC state, and reports the result. ## Quick Start Ask the assistant to validate a specific command or action plan in the lab, for example by saying "vm-lab: run these troubleshooting steps in the lab and show me the verification report."