A production troubleshooting skill in Microsoft's azure-skills: The agent follows a five-step process—"symptoms → resource health → logs → metrics → recent changes"—to diagnose issues with Container Apps, App Service, Function Apps, AKS, VM connectivity, and messaging services on Azure. It can use AppLens, Azure Monitor MCP tools, or az commands for evidence gathering.
skills.sh install count 572729 (+1686 / 24h)
Security Audit: Not audited
- Not yet scanned
Rationale
- Troubleshooting guides are organized by service: image pulls, cold starts, health probes, port mismatches, function invocation failures and binding errors, VM RDP/SSH connectivity issues and NSG blocks, AMQP errors for Event Hubs/Service Bus, etc. AKS events are directed to separate documentation.
- Supports evidence-only mode: When users request or prohibit access to production environments, the agent diagnoses solely based on provided logs, metrics, and timelines, without querying or modifying Azure.
- Includes a library of KQL queries and Azure Resource Graph queries; when resource identifiers are available, it prioritizes using the skill's built-in diagnostic scripts to retrieve data.
- The same repository contains 27 other skills; skills.sh shows 1686 installs in the last 24 hours, with a cumulative total of 572729 (indicating popularity only).
Usage
- Requires an Azure subscription, az CLI login, and read permissions for the relevant resources; using AppLens/Monitor requires connecting to Azure MCP.
- Triggered by phrases like "troubleshoot app service" or "container app 503"; if you do not want it to touch production environments, specify in the prompt that it should perform evidence analysis only.
Risks
- Not audited in this release, assigned a fixed C-level rating, and install commands are not displayed; the built-in diagnostic scripts were not read or scanned.
- The beginning of SKILL.md claims to be an "authoritative guide, must be followed," using strong language that may override your own troubleshooting habits.
- The troubleshooting scope includes operations that modify production resources, such as resetting VM credentials; manual confirmation is required before execution.