Menu
Use when a plan exists and the application source or IaC needs to be created or updated.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Use when starting or refining a project's requirements, before any planning or coding. Supports sub-commands `sync`, `reverse`, and `validate`.
Use when local verification passed and the feature is ready to ship to Azure.
Use when a spec exists and an implementation or Azure deployment plan is needed, before writing code.
Use when the user wants to run the full Specify → Plan → Implement → Deploy loop end-to-end in one shot. Accepts a free-form prompt that is passed to specify.
Use when implementation exists and needs to be exercised locally against provisioned Azure dependencies, before deploying.
Create Azure architecture diagrams, flowcharts, data-flow visualizations, or illustrations as Excalidraw documents with embedded official Azure service icons, and export them as SVG and PNG. Use whenever the user wants to draw, sketch, diagram, visualize, or illustrate an Azure workload, solution, reference architecture, network topology, AI/ML topology, or any cloud system involving Azure services — even if they don't say "Excalidraw". Triggers include: "diagram", "architecture diagram", "draw", "sketch", "visualize", "flowchart", "data flow", "topology", "illustrate" combined with any Azure / cloud context. Also use to re-render existing `.excalidraw` files to SVG/PNG.
| name | implement |
| description | Use when a plan exists and the application source or IaC needs to be created or updated. |
Requires spec.md. If spec.md is missing, auto-run specify with the user's prompt and then auto-run plan. If plan.md is missing, auto-run plan. Load workspace context per copilot-instructions.md.
./infra/ is missing)In an OS temp dir (not the workspace), run one of:
azd init -t <AZD template> -e <AZD environment> -s <Azure Subscription Id> -l <Azure Region>
azd init --minimal -e <AZD environment> -s <Azure Subscription Id> -l <Azure Region>
Pick the form recorded in ./.azure/deployment-plan.md. Treat non-zero exit as failure only if azure.yaml or (non-minimal) infra/ are missing.
Move azure.yaml and infra/ into the workspace root. Ask before overwriting; Delete the temp dir.
./.azure/deployment-plan.md must have no [NEEDS CLARIFICATION: …] markers. Run azd env set for AZURE_RESOURCE_GROUP first, then for every other variable in the deployment plan.
Implement per ./docs/plan.md, keeping it updated with progress. On resume, re-read ./docs/plan.md and continue from the next unchecked step.
Accelerate with parallelism. Keep shared contracts (APIs, env vars, azure.yaml service names) consistent across agents, and serialize steps with true dependencies (scaffold and env-var persistence before service code).
Create or update ./docs/implementation.md to capture the architecture that was actually built and the key implementation details (components, data flow, key decisions, deviations from plan.md). Include one or more animated Mermaid flowchart diagrams to visualize the runtime architecture and request/data flow.
Pause if: a task is unclear, implementation reveals a spec/plan gap, or any error/blocker is hit — report and wait, do not guess.
Check the following and fix any gaps before claiming implementation is complete and ready for verify/deploy:
./azure.yaml has service configuration (host: azure.ai.agent) for each implemented agent../azure.yaml has service configuration for each MCP server../azure.yaml has service configuration for the frontend../azure.yaml has service configuration for the backend../azure.yaml has services configured with cloud build (remoteBuild: true) set for container images. Use docker if the user prefers local build and it's running; otherwise, use ACR cloud build../infra).azd ai agent show <agent-name> and reading the Endpoint protocol and URL, and authenticate with a DefaultAzureCredential bearer token (scope https://ai.azure.com/.default). Make sure to use the protocol supported by the agent (invocations, responses, etc.)../infra provision its own registry; container apps need it even though hosted agents use Foundry's remote build.Summarize the updated ./docs/implementation.md (architecture, key decisions, and any deviations from plan.md), report the result of the post-implementation checks, and prompt the user to continue with verify (local tests against provisioned Azure resources) or skip straight to deploy (provision and deploy to Azure with azd up).