Source: microsoft/azure-skills · Maintainer: Microsoft · License: MIT · Checked: 27 October 2026
What is the Azure compute skill?
The Azure compute skill helps a coding agent choose, price and create Azure virtual machines and VM Scale Sets (VMSS). It asks about your workload, suggests two or three sizes with live prices and quota status, then writes az CLI, Bicep or Terraform, or creates the VM through the Azure MCP Server after you approve a plan. It is for developers who need a bare server or scale set, not a full app deployment.
The SKILL.md calls itself an "Azure VM/VMSS router", and its opening line says it "Routes Azure VM and Virtual Machine Scale Set (VMSS) requests to the right workflow." The trigger list covers creating VMs, recommending sizes, comparing pricing, scale sets, autoscale, GPU and HPC workloads, capacity reservations and machine enrollment.
The skill ships in microsoft/azure-skills, the Azure Skills Plugin, where the README lists it under "Optimize architecture and cost". The frontmatter lists version 2.5.1 and author Microsoft.
As of 27 October 2026, its skills.sh page listed about 523,100 installs (snapshot, 27 October 2026).
What does it do?
The SKILL.md is a short router, about 2.6 KB. It tells the agent to classify the request, open one workflow file and load only the references that workflow asks for. The folder holds about 40 files in all.
| Workflow | Use when |
|---|---|
| VM Recommender | Choosing a VM or VMSS, autoscaling, pricing or comparing options |
| VM Creator | Creating a bare VM or VMSS |
| Capacity Reservation | Reserving VM capacity with a Capacity Reservation Group |
| Essential Machine Management | Enrolling a subscription's machines in management |
VM Recommender. It gathers requirements (workload, vCPU and RAM, GPU, OS, region, scaling, availability, load balancing), decides VM versus VMSS with a short table, picks two or three VM families, and verifies them against live Microsoft Learn pages. Prices come from the Azure Retail Prices API, which the workflow says needs no Azure subscription. Before showing results it checks quota if you have a subscription, then presents options with hosting model, size, specs, instance count, estimated hourly price, why it fits and the trade-off. For scale sets it notes that VMSS has no extra charge and that pricing is per VM.
VM Creator. It adapts its questions to you: a beginner path asks at most two (region, then Linux or Windows) and fills in defaults such as Standard_D2s_v5, Ubuntu 24.04, a new virtual network, and an NSG that allows SSH or RDP only from your current public IP. Expert paths go deeper on networking, specs, cost or security. Read-only checks confirm the SKU, image, quota and region before it renders a Plan Card, a table of every decision for you to approve or edit. Then you pick the output:
| Output | What you get |
|---|---|
| az CLI | A bash script to run yourself |
| Bicep | An infrastructure-as-code template |
| Terraform | Terraform configuration (the folder includes a main.tf example) |
| Apply via Azure MCP | A live deployment, confirmed once more first |
File outputs can be printed, saved locally or opened as a GitHub pull request with the gh CLI. The PR mode refuses to commit straight to main.
Capacity Reservation. It creates Capacity Reservation Groups with az capacity reservation group create and associates or disassociates VMs and scale sets. It always asks how many instances to reserve and warns that charges start when the reservation is created, whether or not VMs use it. Microsoft's capacity reservation overview has the full billing rules.
Essential Machine Management. It enables, checks and offboards subscription-level machine management, which the workflow marks as "currently in public preview."
How do you install the Azure compute skill?
The repo README documents several routes. The full plugin fits best, because quota checks, SKU lookups and the apply step use Azure MCP Server tools such as compute_vm_list-skus and compute_vm_check-quota.
Claude Code: the Azure Skills Plugin. Run inside Claude Code:
/plugin install azure@claude-plugins-official
Or run /plugin and search for "azure". Update later with /plugin update azure@claude-plugins-official. See Claude Code plugins and Claude Code MCP.
GitHub Copilot CLI:
/plugin marketplace add microsoft/azure-skills
/plugin install azure@azure-skills
Codex, Cursor and Gemini CLI. For Codex, the README says to run codex plugin marketplace add microsoft/azure-skills, then install azure from /plugins. For Cursor, use the Cursor Marketplace or Settings, then Plugins. For Gemini CLI, run gemini extensions install https://github.com/microsoft/azure-skills. See Codex skills and Cursor skills.
Skill only. The command on skills.sh copies just this folder:
npx skills add https://github.com/microsoft/azure-skills --skill azure-compute
Without Azure MCP, the creator warns that quota and SKU availability are unverified and offers the CLI route.
Prerequisites, per the README: an Azure account or subscription, Node.js 18 or later on your PATH for npx, and the Azure CLI signed in with az login.
How do you use it?
Describe the workload and constraints. The skill is model-invoked, so phrases like "which VM size", "spin up a VM" or "scale set" load it.
Example prompts (suggestions, not quotes from the repo):
Which Azure VM should I use for a Postgres server with 4 cores and 32 GB RAM in westeurope? Show prices.Compare a single VM and a scale set for a stateless API that peaks at 10x traffic on Mondays.Create an Ubuntu VM in eastus for a dev box. Give me the Bicep and save it to infra/devbox.Just print the az CLI for a Standard_D4s_v5 VM, no questions.Reserve capacity for 6 Standard_E8s_v5 VMs in eastus before our launch.
The fourth prompt uses the creator's fast path: when you name a deliverable and refuse questions, it shows a one-line preview and emits the artifact, though the validation checks still run.
Tips and limits
From the repo, Microsoft Learn and skills.sh, not from testing:
- Read the Plan Card. It lists silent defaults, including the NSG source. The beginner path falls back to allowing any source (
*) only if detecting your IP fails, and flags it. - It reads local files and calls a third party. The beginner path reads
~/.ssh/id_rsa.pubfor the SSH key and detects your IP withcurl -s ifconfig.me. - Region defaults to eastus. If you don't name one, the recommender uses
eastusand notes that prices vary by region. - Reservations bill immediately. Only reserve what you'll use.
- Security audits. On skills.sh, Snyk rated the skill Warn (medium risk) on 15 September 2026, flagging third-party content exposure from the optional GitHub PR mode, where the agent may read PR, issue or comment text written by outsiders. Gen Agent Trust Hub rated it Pass (Safe) on 15 September 2026, noting the
ifconfig.mecall, the SSH public key read, runtimeaz extension addinstalls andghrepo operations. Socket rated it Pass on 15 September 2026. - Telemetry. The README says to set
AZURE_MCP_COLLECT_TELEMETRY=falseto turn off Azure MCP telemetry.
The MIT License lets you copy and change the skill as long as you keep Microsoft's copyright and license notices.
Save it in your Kipt library
An edited copy pays off if your team has standard answers: an approved region, a default VM family, required tags, or "Terraform only." Put them in the defaults and the agent stops asking. In Kipt, open Skills and paste the skill's GitHub folder link into the import box, or drop the folder or a .zip; Kipt keeps the workflow and reference files with it. The copy does not include the plugin's MCP setup. Skills in your private GitHub repo install into Claude Code or Codex with one command:
npx skills add https://github.com/<you>/<repo>.git --all -g
Kipt is not affiliated with Microsoft.
Similar skills
- Azure prepare skill: the skill azure-compute hands off to when you're deploying an application, not a bare VM.
- Azure validate skill: preflight checks before an Azure deployment.
- Azure deploy skill: runs azd and Terraform deployments.
- Terraform skill: HashiCorp's skill, useful if you take the Terraform output and maintain it yourself.
- Azure compliance skill: scans the resources you create against Azure best practices.