Analyze Workspace

⛔ MANDATORY FIRST — Specialized Technology Delegation

STOP. Before choosing a mode, check the user's prompt for specialized technology keywords. If matched, invoke the corresponding skill immediately — it has tested templates and correct SDK usage.

⚠️ Re-entry guard: If azure-prepare was invoked as a resume from a specialized skill (e.g., azure-hosted-copilot-sdk Step 4), skip this check and go directly to Step 4.

⚠️ Check the user's prompt text — not just existing code. This is critical for greenfield projects with no codebase. See full routing table.

If no match, continue below.

Three Modes — Always Choose One

⛔ IMPORTANT: Always go through one of these three paths. Having azure.yaml does NOT mean you skip to validate — the user may want to modify or extend the app.

Decision Tree

What does the user want to do?
│
├── Create new application → Mode: NEW
│
├── Add/change features to existing app
│   ├── Has azure.yaml/infra? → Mode: MODIFY
│   └── No Azure config? → Mode: MODERNIZE (add Azure support first)
│
└── Migrate/modernize for Azure
    ├── Cross-cloud migration (AWS/GCP/Lambda)? → **Invoke azure-cloud-migrate skill** (do NOT continue in azure-prepare)
    └── On-prem or generic modernization → Mode: MODERNIZE

Mode: NEW

Creating a new Azure application from scratch.

Actions:

1. Confirm project type with user
2. Gather requirements → requirements.md
3. Select technology stack
4. Update plan

Mode: MODIFY

Adding components/services to an existing Azure application.

Actions:

1. Scan existing codebase → scan.md
2. Identify existing Azure configuration
3. Gather requirements for new components
4. Update plan

Mode: MODERNIZE

Converting an existing application to run on Azure.

Actions:

1. Full codebase scan → scan.md
2. Analyze existing infrastructure (Docker, CI/CD, etc.)
3. Gather requirements → requirements.md
4. Map existing components to Azure services
5. Update plan

Detection Signals

⛔ MANDATORY for Azure Functions: Load Composition Rules BEFORE Execution

If the target compute is Azure Functions, you MUST load the composition algorithm before generating ANY infrastructure:

1. Load services/functions/templates/selection.md — decision tree for base template + recipe
2. Load services/functions/templates/recipes/composition.md — the exact algorithm to follow
3. Use functions_template_get MCP tool to list and fetch templates and write functionFiles[] + projectFiles[] directly — NEVER hand-write Bicep/Terraform and use azd init -t <template>/func init/func new as fallback when composing multiple recipes and required templates are not found

⚠️ Critical: The Functions bicep.md and terraform.md files are REFERENCE DOCUMENTATION, not templates to copy. Hand-writing infrastructure from these patterns results in missing RBAC, incorrect managed identity configuration, and security vulnerabilities.

For other compute targets (Container Apps, App Service, Static Web Apps), load their respective README files in services/ for guidance.