azure-prepare
Assess an application and prepare its project configuration for Azure deployment.
npx skills add microsoft/azure-skills --skill azure-preparePrepare, validate and troubleshoot Azure application deployments.
Begin with project preparation for a new service. Use diagnostics when a deployed service already has an error to reproduce.
Assess an application and prepare its project configuration for Azure deployment.
npx skills add microsoft/azure-skills --skill azure-prepareCheck Azure project configuration and deployment prerequisites.
npx skills add microsoft/azure-skills --skill azure-validateDeploy a prepared application to the intended Azure environment.
npx skills add microsoft/azure-skills --skill azure-deployInvestigate Azure runtime issues using resource state, logs and platform diagnostics.
npx skills add microsoft/azure-skills --skill azure-diagnosticsChoose and configure Azure storage for application data and files.
npx skills add microsoft/azure-skills --skill azure-storageWork with Microsoft Foundry projects, models and AI application resources.
npx skills add microsoft/azure-skills --skill microsoft-foundryNo skill in this category has a full write-up yet.
A cloud skill turns “deploy this” from a conversation into an operation. The model already knows what a Dockerfile looks like; what it does not know is your subscription, your region, your resource names, and the order in which your platform accepts changes. A cloud skill supplies exactly that — the account-specific steps and a check after each one.
The honest promise: it saves you from re-explaining your setup on every run, and it stops the agent from inventing flags that your CLI version no longer has.
All six skills on this page come from microsoft/azure-skills, and they are a sequence rather than alternatives:
azure-prepare — reads the application and writes the project's deployment configuration: resource names, region, app settings. This is where the deployment target is decided. Skip it and the agent guesses.azure-validate — checks that configuration against the prerequisites before anything is created. Cheap failure goes here.azure-deploy — creates and updates the resources. Expensive failure goes here, which is why validate exists.azure-diagnostics — reads resource state and logs when something answers 403 or 404 after a successful deploy. That pairing — deploy succeeded, first request failed — is almost always identity and role assignment, not a code bug.azure-storage and microsoft-foundry — the two sideways skills. Storage decides where application data and files live; Foundry is for Microsoft's AI project resources.If you install only azure-deploy, expect it to guess the target. That is the single most common way an Azure run goes wrong.
--help before running.Database skills matter the moment a deploy touches migrations — a deploy that succeeds but leaves the schema one version behind is worse than a failed one. Agent workflow explains the SKILL.md format these skills are written in.
Review the target subscription and environment, then validate configuration and service health.
Confirm the resource group and region in the portal match the intended environment, then load the health endpoint and read the start-up log. A deploy that returns 403 or 404 on the first request usually means identity and role assignment, not a code bug.
Use the command beside the skill, select a supported agent, then reload the agent and confirm it can see the skill in your project.
You can install azure-deploy on its own. Before deploying, run azure-prepare to set the resource names, region and app settings, then use azure-validate to verify the target.
No. The skills in microsoft/azure-skills describe Azure resources and Azure deployment flows. For AWS or GCP you need that platform's own skills.
Work in a separate resource group or subscription, run validate before deploy, and keep the previous deployment available so you can switch back without rebuilding configuration from memory.
Check that the calling identity has the required Azure role. Use azure-diagnostics to inspect resource state and logs before you change configuration.
Only as current as the skill's last update. Treat unfamiliar flags as a signal to verify before running the command.