Azure Architecture Autopilot

Azure Architecture Builder

A pipeline that designs Azure infrastructure using natural language, or analyzes existing resources to visualize architecture and proceed through modification and deployment.

The diagram engine is embedded within the skill (scripts/ folder). No pip install needed — it directly uses the bundled Python scripts to generate interactive HTML diagrams with 605+ official Azure icons. Ready to use immediately without network access or package installation.

Automatic User Language Detection

🚨 Detect the language of the user’s first message and provide all subsequent responses in that language. This is the highest-priority principle.

• If the user writes in Korean → respond in Korean
• If the user writes in English → respond in English (ask_user, progress updates, reports, Bicep comments — all in English)
• The instructions and examples in this document are written in English, and all user-facing output must match the user’s language

⚠️ Do not copy examples from this document verbatim to the user. Use only the structure as reference, and adapt text to the user’s language.

Tool Usage Guide (GHCP Environment)

All sub-agents (explore/task/general-purpose) cannot use web_fetch or web_search. Fact-checking that requires MS Docs lookups must be performed directly by the main agent.

External Tool Path Discovery

az, python, bicep, etc. are often not on PATH. Discover once before starting a Phase and cache the result. Do not re-discover every time.

⚠️ Do not use Get-Command python — risk of Windows Store alias. Direct filesystem discovery ($env:LOCALAPPDATA\Programs\Python) takes priority.

az CLI path:

$azCmd = $null
if (Get-Command az -ErrorAction SilentlyContinue) { $azCmd = 'az' }
if (-not $azCmd) {
  $azExe = Get-ChildItem -Path "$env:ProgramFiles\Microsoft SDKs\Azure\CLI2\wbin", "$env:LOCALAPPDATA\Programs\Azure CLI\wbin" -Filter "az.cmd" -ErrorAction SilentlyContinue | Select-Object -First 1 -ExpandProperty FullName
  if ($azExe) { $azCmd = $azExe }
}

Python path + embedded diagram engine: refer to the diagram generation section in references/phase1-advisor.md.

Progress Updates Required

Use blockquote + emoji + bold format:

> **⏳ [Action]** — [Reason]
> **✅ [Complete]** — [Result]
> **⚠️ [Warning]** — [Details]
> **❌ [Failed]** — [Cause]

Parallel Preload Principle

While waiting for user input via ask_user, preload information needed for the next step in parallel.

Path Branching — Automatically Determined by User Request

Path A: New Design (New Build)

Trigger: “create”, “set up”, “deploy”, “build”, etc.

Phase 1 (references/phase1-advisor.md) — Interactive architecture design + diagram
    ↓
Phase 2 (references/bicep-generator.md) — Bicep code generation
    ↓
Phase 3 (references/bicep-reviewer.md) — Code review + compilation verification
    ↓
Phase 4 (references/phase4-deployer.md) — validate → what-if → deploy

Path B: Existing Analysis + Modification (Analyze & Modify)

Trigger: “analyze”, “current resources”, “scan”, “draw a diagram”, “show my infrastructure”, etc.

Phase 0 (references/phase0-scanner.md) — Existing resource scan + diagram
    ↓
Modification conversation — "What would you like to change here?" (natural language modification request → follow-up questions)
    ↓
Phase 1 (references/phase1-advisor.md) — Confirm modifications + update diagram
    ↓
Phase 2~4 — Same as above

When Path Determination Is Ambiguous

Ask the user directly:

ask_user({
  question: "What would you like to do?",
  choices: [
    "Design a new Azure architecture (Recommended)",
    "Analyze + modify existing Azure resources"
  ]
})

Phase Transition Rules

• Each Phase reads and follows the instructions in its corresponding references/*.md file
• When transitioning between Phases, always inform the user about the next step
• Do not skip Phases (especially the what-if between Phase 3 → Phase 4)
• 🚨 Required condition for Phase 1 → Phase 2 transition: 01_arch_diagram_draft.html must have been generated using the embedded diagram engine and shown to the user. Do not proceed to Bicep generation without a diagram. Completing spec collection alone does not mean Phase 1 is done — Phase 1 includes diagram generation + user confirmation.
• Modification request after deployment → return to Phase 1, not Phase 0 (Delta Confirmation Rule)

Service Coverage & Fallback

Optimized Services

Microsoft Foundry, Azure OpenAI, AI Search, ADLS Gen2, Key Vault, Microsoft Fabric, Azure Data Factory, VNet/Private Endpoint, AML/AI Hub

Other Azure Services

All supported — MS Docs are automatically consulted to generate at the same quality standard. Do not send messages that cause user anxiety such as “out of scope” or “best-effort”.

Stable vs Dynamic Information Handling

Quick Reference