How to use the Azure Resource Health Diagnosis agent skill

What the skill makes the agent do

The skill is a file named SKILL.md that describes a step-by-step workflow. Once it is installed, the agent follows that workflow when you ask it to look into an Azure resource such as a web app, database, virtual machine or storage account.

The workflow checks the resource's status and metrics, runs queries against its logs and telemetry to find errors and patterns, and sorts the problems it finds by severity. It then writes a remediation plan with immediate, short-term and longer-term actions.

• Resource lookup: finds the resource by name, and asks you to specify subscription or resource group if several match.
• Health check: reviews provisioning state, availability, recent changes and utilization.
• Log analysis: finds Log Analytics workspaces and Application Insights instances, then runs KQL queries.
• Classification: ranks issues as Critical, High, Medium or Low.
• Remediation plan: lists immediate, short-term and long-term actions, with rollback plans and verification steps, and asks for your approval before acting.

What you need before you start

The skill depends on Azure's own tools. You need an Azure MCP server that is configured and authenticated, and the Azure CLI (az) is used as a fallback when needed. You also need access to an Azure subscription that contains a deployed, running target resource with its Log Analytics or Application Insights data.

Without a subscription or access to the resource, the workflow cannot run.

• Azure MCP server configured and authenticated
• Azure CLI (az) as a fallback
• An Azure subscription with a deployed, running target resource
• Log Analytics and/or Application Insights data for that resource

Install the skill from the ZIP download

Download the ZIP for this skill from this page and unzip it. The package contains a single file, SKILL.md. Place the unzipped skill folder in one of two locations depending on how you want to use it.

For a personal install available across your projects, put it in ~/.claude/skills/. For a single project, put it in that project's .claude/skills/ folder.

In the Claude apps you can instead upload the skill under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/
• Project only: .claude/skills/ inside the project folder
• Claude apps: Settings → Capabilities → Skills, then upload

Trigger the skill with natural requests

You do not need a special command. Describe the problem and name the resource, and the agent should pick up the skill from its description. Include the resource name, and the resource group or subscription if you know them, so the lookup is quicker.

Requests that work well describe a symptom and a target, for example asking the agent to check why a named web app is returning errors, or to review a database that has been slow since yesterday.

• "Diagnose the health of the web app orders-api in resource group prod-rg."
• "Why is the storage account logsarchive showing failed requests this week? Make a remediation plan."
• "Check my SQL database for connection failures in the last 24 hours."

Read the output and verify before acting

The skill asks for confirmation before remediation work. The health summary lists the resource, its status, location and the number of issues at each severity level, followed by the plan.

Treat the plan as a well-organized starting point. The skill's own guidance says the plan is only as good as the available logs and the agent's interpretation, so a person should verify findings before changing anything in production. Remediation steps can affect live services.

• Check that the cited log results match what you see in the portal.
• Review each rollback plan before applying a fix.
• Test changes outside production where you can.

Safety notes: scripts and network access

This skill does not run scripts. It is made of instructions only. It does use the network, because it calls Azure services through the Azure MCP tools and the Azure CLI to read resource status, metrics and logs. Those calls run under the credentials you have authenticated with.

Use an account with the least access needed to read the resource and its logs. Do not apply remediation commands from the plan to production without review.

• Runs scripts: no
• Uses network: yes, to query Azure
• Actions run under your authenticated Azure identity