What does this skill do, and when should you use it?
Azure Compute is a routing skill for Azure virtual machines and Virtual Machine Scale Sets. It covers sizing and pricing comparisons, bare VM/VMSS creation and deployment, RDP and SSH troubleshooting, Capacity Reservation Groups, and Essential Machine Management. It requires workflow-first routing before supporting references are loaded. It is intended for bare infrastructure; application deployments belong with `azure-prepare`.
Classifies requests into VM Recommender, VM Creator, VM Troubleshooter, Capacity Reservation, or Essential Machine Management workflows; prioritizes the Azure MCP tools compute_vm_list-skus, compute_vm_list-images, and compute_vm_check-quota for VM creation intents; and handles VM sizing, pricing, VMSS, autoscaling, connectivity failures, password resets, NSGs, CRGs, and EMM requests.
- An infrastructure team choosing Azure VM sizes for a website backend, GPU workload, machine-learning job, or HPC simulation.
- An engineer provisioning or deploying a bare Azure VM or VMSS.
- An operator diagnosing refused RDP or SSH connections, port 3389 issues, black screens, or password problems.
- An Azure user reserving capacity or creating, associating, or disassociating a Capacity Reservation Group.
- An administrator asking about Essential Machine Management enrollment or monitoring.
How do you install this skill?
- Deployment, updates, deletion, CRG association/disassociation, and EMM enablement can create real Azure resources, cause downtime, or incur ongoing charges; show scope, permissions, impact, cost, and rollback steps and obtain confirmation before execution.
- The VM Creator, VM Troubleshooter, and several EMM/workflow references are not supplied, so complete workflows, command compatibility, and failure feedback cannot be confirmed.
- Prices, SKUs, quotas, VMSS limits, and preview APIs can change; verify current official documentation and subscription/region state before giving actionable advice.
- The material does not demonstrate Chinese-language quality or mainland-China reachability; separately check network access and authentication for Azure, Microsoft Learn, GitHub Copilot, MCP, and external API dependencies.
- Network access
- Local filesystem
- MCP Server
Azure MCP servercompute_vm_list-skuscompute_vm_list-imagescompute_vm_check-quota
Run npx skills add microsoft/skills in the project and select azure-compute in the installation wizard. Skills are installed in the selected agent directory, such as .github/skills/ for GitHub Copilot.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/microsoft/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/.github/plugins/azure-skills/skills/azure-compute" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Compare Azure VM sizes and pricing for a machine-learning workload
- My Azure VM refuses SSH connections.
Use a concrete request such as “Compare Azure VM sizes and pricing for a machine-learning workload” or “My Azure VM refuses SSH connections.” For Docker services, web apps, APIs, or serverless workloads, use azure-prepare instead.
What are this skill's strengths and limitations?
- Covers VM and VMSS selection, autoscaling, pricing comparison, connectivity troubleshooting, CRGs, and EMM.
- Defines explicit workflow-first routing and warns against entering through reference files.
- The provided frontmatter identifies Microsoft as author, MIT as license, and version 2.4.4.
- It is limited to bare VM/VMSS infrastructure and does not handle application deployment.
- The supplied material does not include the detailed contents of the workflow files, permission requirements, or complete troubleshooting procedures.
- It depends on Azure MCP tools; the supplied material does not establish specific integrations with other agent platforms.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Azure Compute Navigator this page ✓ Microsoft · Official | 50 · Use with care | ★ 2.7k | 3mo ago | MIT |
| AKS Automatic Readiness ✓ Microsoft · Official | 55 · Use with care | ★ 2.7k | 3mo ago | MIT |
| Azure Enterprise Infrastructure Planner ✓ Microsoft · Official | 48 · Use with care | ★ 2.7k | 3mo ago | MIT |
| Google Cloud Performance Advisor ✓ Google · Official | 50 · Use with care | ★ 21k | 11d ago | Apache-2.0 |
| GKE AI Inference Deployment Assistant ✓ Google · Official | 46 · Use with care | ★ 21k | 11d ago | Apache-2.0 |
The repository distinguishes it from azure-prepare: Azure Compute handles bare VM/VMSS infrastructure, while azure-prepare handles Docker services, web apps, APIs, and serverless application deployments.
How did FollowSkills review this skill?
The evidence shows workflow-first routing, disclosure of Azure CLI, the unauthenticated Retail Prices API, and Azure management APIs, plus warnings about billing, permissions, and preview status for CRG/EMM. MIT licensing, Microsoft provenance, and some command scoping are traceable. Points are deducted because uniform least-privilege requirements, pre-execution confirmation, sensitive-data handling, network data-flow disclosure, and rollback/recovery are not specified; several commands create, update, delete, or alter subscription-level resources.
The evidence provides routing tables, requirement collection, two-level quota checks, pagination guidance, error cases, and some abnormal-input handling. Points are deducted because this review did not execute anything, and the selected skill references unprovided VM Creator, VM Troubleshooter, and several EMM/workflow reference files. Some paths depend on CLI extensions, MCP, authentication, or preview APIs, so key-path execution cannot be confirmed statically; the score stays below the static cap of 10.
The evidence covers VM/VMSS recommendation, deployment, connectivity failures, quotas, CRGs, and EMM, with explicit boundaries against azure-prepare and vm-creator and a broad trigger vocabulary. Points are deducted because the scope is broad, several capabilities require an authenticated Azure environment and external tools, and Chinese interaction, unavailable-environment behavior, non-Azure exclusions, fine-grained false-trigger boundaries, and mainland-China reachability are not documented.
The evidence includes YAML metadata, MIT licensing, version 2.4.4, workflow-first routing, progressive reference loading, command examples, limitations, and repository-level governance signals from README, CI, and SECURITY.md. Points are deducted for the absence of a skill-specific changelog, explicit maintainer/update path, FAQ, installation prerequisites, and a complete reference inventory; the README counts also differ from the stated repository count, and several linked files are absent from the supplied material.
The evidence shows useful intent-to-workflow mapping and reusable guidance for quota calculation, price lookup, VMSS selection, and CRG operations. Points are deducted because the core deployment and troubleshooting workflows are missing from the supplied evidence, many recommendations require live verification, and there is no execution evidence for correctness, completeness, direct usability, or marginal benefit over manual alternatives; the static rule limits this to 7.
The evidence includes revision-scoped source text, commands, error tables, CI workflows, and committed harness tests, providing limited auditability. Points are deducted because the tests do not demonstrate coverage of this azure-compute skill's key paths, nothing was executed or independently reproduced, external links were not verified in this review, and some claims are only self-authored skill assertions; the static rule limits this to 5.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →