A workspace skill for proceeding with verification, PoC, construction, and confirmation involving configuration changes in an Azure environment under the premise of a safe lab/sandbox.
Read-only triage of production incidents is out of scope; this skill covers creation of verification configurations, state monitoring, constraint workarounds, before/after comparison, cleanup up to these.
When to Use
- When you want to newly build a verification environment on Azure and try out features and constraints
- When you want to test networking features in a lab such as VPN Gateway, VNet peering, BGP, Route Server, Private Endpoint, and Hub-and-Spoke
- When you want to assemble a minimal Azure-only verification as an alternative to a production ExpressRoute configuration
- When you want to proceed with read/write infrastructure verification using Azure CLI
- When you want to monitor Azure resources during deployment and move to the next step once they become READY
- When you want to move forward while working around Azure SKU / zone / auth / region constraints that arise during verification on the spot
- When you want to reflect the construction results and before/after observed values in
Answer/やreports/when you want to reflect it in
When NOT to Use
- When you want to triage a production incident in read-only mode
- When you only want simple product Q&A or specification confirmation
- → Normal Docs research
- When you want to complete end-to-end verification of a production ExpressRoute that requires a carrier or on-premises hardware here and now
- With this skill, approximate verification in a lab is possible, but provider-side peering and actual hardware FIC are not covered
- When you want to design and deploy a configuration intended for reuse with Bicep / AVM
Inputs
At minimum, receive two or more of the following.
- tenant ID
- subscription ID
- The name of the feature you want to verify
- The assumed topology (e.g., hub-and-spoke, VPN BGP, private endpoint)
- Cost tolerance
- Whether cleanup is required
If anything is missing,どの tenant/subscription に作るか と 検証したい到達点 confirm with top priority.
Core Rules
- First, whether you will touch production or use a lab/sandbox make this clear
- If touching production, do not make destructive changes. For verification, prioritize a separate subscription
- For verification involving changes,minimal configuration / low-cost configuration start from
- Do not build the full configuration right away;the minimal unit necessary to achieve the objective divide by this and then expand
- Fix the Azure CLI authentication state, tenant, and subscription first
- Deploy after checking prerequisites and constraints in Microsoft Learn
- Before deployment,whether the verification truly holds with Azure-only / whether actual hardware or a provider is required perform a feasibility check
- For long-running Azure operations during deployment,state monitoring script use this or status polling to proceed to the next step
- In verification of configuration changes or setting changes,prepare an observation script before the changeand capture the control-plane change time and the dataplane / route impact time separately
- consider control-plane validation and data-plane / route validation separately
- always leave a before/after comparison
あとで Activity Log を見ればよいwithout assuming that, use polling immediately after the change to capture reconvergence time and whether there was a brief outage- after validation, clearly state the cleanup policy
Decision Points
1. Is ExpressRoute itself necessary?
- if what is needed is only functional validation of route advertisement , then prioritize an Azure-only lab based on VPN Gateway
- if validation requires active circuit / private peering / BGP peering / provider-side provisioning , then explicitly state that this skill cannot "fully substitute" for production ExpressRoute validation
2. Is Azure-only minimal validation sufficient?
- Sufficient: VNet-to-VNet VPN + BGP + hub/spoke +
summarizedGatewayPrefixes - Insufficient: you want to see the quirks of on-prem physical peer or FIC-specific constraints
- → In a second phase, extend to pseudo on-prem using NVA / FRR / strongSwan / Bird
2.5. Pass the feasibility check first
Before starting, briefly determine the following.
- The purpose is functional feasibility confirmation か、production-equivalent end-to-end validation か
- Can it be substituted with Azure-only?
- Can you start with a low-cost SKU / minimal number of spokes?
- Can you assume you will increase them along the way?
If this determination shows Azure-only is sufficient, then at first proceed with minimal number of spokes / minimal number of gateways / minimal number of routes .
For deployment constraints, READY determination, and detailed phase procedures, refer to references/validation-runbook.md .
Workflow
- Feasibility: decide the validation goal, Azure-only feasibility, minimal configuration, cost/cleanup
- Scope: fix tenant / subscription / lab or production boundary / target endpoint
- Official grounding: check prerequisites / limits / supported SKUs in Microsoft Learn
- Preflight:
az account show, check provider registration, region availability, RBAC, and observation methods - Build: create the minimal Resource Group / VNet / GatewaySubnet / peering / core resources
- Observe:
READY/Connectedwait until , then save baseline routes / peer status / metrics - Change: change only one target setting, and record the requested time, Succeeded time, and correlation ID
- Compare: evaluate before/after, reconvergence time, and whether there was a brief outage using polling evidence
- Cleanup or persist: keep or delete the lab, and reflect the results in the report / Answer
Detailed steps: references/validation-runbook.md
Done Criteria
- The target tenant / subscription is fixed
- The validation topology is a minimal configuration
- SKU / zone / address plan have been adjusted according to Azure constraints
-
READY/Connectedyou have confirmed before proceeding to the next process - before/after route information remains
- any of the change request time, Succeeded time, or impact observation time remains
- The reconvergence time or presence/absence of momentary interruptions can be explained not only by the Activity Log but also by polling evidence
- The results were
Answer/やreports/reflected in - left a cleanup policy
Output Shape
Include in the report 検証目的、検証構成、事前状態、変更内容、変更時刻、事後状態、評価、Cleanup The template uses references/output-template.md .
Example Prompts
/azure-infra-validation <lab-subscription> で VPN Gateway ベースの summarizedGatewayPrefixes 検証ラボを作って/azure-infra-validation Azure-only で route advertisement の before/after を見たい。最小構成を作って/azure-infra-validation hub-and-spoke + BGP の lab を構築し、設定変更後の learned routes を比較して/azure-infra-validation Azure CLI だけで構築可能な検証プランを作って、そのまま実行して/azure-infra-validation デプロイ中の Gateway を監視し、READY になったら次工程へ進めて
Integration Notes
- This skill is limited to read-write verification in lab/sandbox, and separates responsibilities from read-only investigation of production incidents
- If the verification results are useful for interpreting a production incident, the investigation results may be passed as input to the production investigation