Architect Phase Skill

ROOT AGENT ONLY - Spawns architect agent, runs only from root Claude Code agent.

Purpose: Enrich user stories with implementation architecture details Trigger: After design-phase completes (stories extracted and PR approved) Input: storiesFolder (string), projectFolder (string), complexity (int), scope (object), tools (string[]) Output: {enrichedStories, architectureNotes, patternReferences}

Workflow

1. Spawn architect agent (enrichment mode)

• Read all story files from storiesFolder
• For each story:
  • Identify story requirements and acceptance criteria
  • Check multi-mono for existing solutions (if applicable)
  • Find example files in reference repos (from scope)
  • Validate patterns against Context7 docs (if external libraries used)
  • Add "Architecture Details" section with:
    • API Endpoints: Routes and methods (e.g., POST /api/users)
    • Files to Create/Modify: Exact paths (e.g., services/auth/routes.ts)
    • Import Statements: Key dependencies needed
    • Database Models: Schemas or ORM definitions
    • Component Names: React/UI component references
    • Patterns: Reference existing patterns from codebase
• Save updated story files
• Return paths to all enriched stories

2. Generate cross-cutting concerns document (if needed)

• If stories have complex interdependencies:
  • Create architecture-notes.md in project folder
  • Document shared patterns, middleware, utilities
  • Add dependency graph or integration points
• Complexity ≥30 → deeper analysis mode

3. Output enriched stories

• Return list of enriched story file paths
• Return architecture notes path (if created)
• Return list of pattern references found

Architecture Annotation Pattern

Story File (before):

# User Story: Admin Dashboard Setup

**As a** system administrator
**I want to** view a dashboard showing system metrics
**So that** I can monitor application health

## Acceptance Criteria

- Dashboard displays CPU usage, memory, and response times
- Data refreshes every 10 seconds
- Show metrics for last 24 hours

Story File (after architect annotation):

# User Story: Admin Dashboard Setup

**As a** system administrator
**I want to** view a dashboard showing system metrics
**So that** I can monitor application health

## Acceptance Criteria

- Dashboard displays CPU usage, memory, and response times
- Data refreshes every 10 seconds
- Show metrics for last 24 hours

## Architecture Details

- **API Endpoints:**
  - `GET /api/admin/metrics` - Returns system metrics
  - `GET /api/admin/metrics/history?range=24h` - Historical data

- **Files to Create/Modify:**
  - `services/admin/routes/metrics.routes.ts`
  - `services/admin/controllers/metrics.controller.ts`
  - `web/src/pages/AdminDashboard.tsx`
  - `web/src/components/MetricsCard.tsx`

- **Import Statements:**
  ```typescript
  import { metricsService } from "@shared/services/metrics";
  import { formatMetrics } from "@shared/utils/formatting";
  ```

• Database Models:

  • MetricsSnapshot - stores timestamp, cpu, memory, responseTime
  • Index on createdAt for range queries

• Component Names:

  • AdminDashboard - main page
  • MetricsCard - metric display card
  • RefreshButton - 10s auto-refresh trigger

• Patterns:

  • Real-time updates: WebSocket with fallback to polling (see rugby-crm for similar pattern)
  • Data formatting: Use @shared/utils/formatting (consistent with existing code)
  • API error handling: Standard error middleware (see services/common/middleware)

---

## Input Specification

```json
{
  "storiesFolder": "docs/projects/msm008-feature/user-stories/",
  "projectFolder": "docs/projects/msm008-feature/",
  "complexity": 26,
  "scope": {
    "targets": ["/home/user/code/metasaver-com"],
    "references": ["/home/user/code/rugby-crm"]
  },
  "tools": ["serena"]
}

Output Specification

{
  "enrichedStories": [
    "user-stories/US-001-database-schema.md",
    "user-stories/US-002-contracts-types.md",
    "user-stories/US-003-workflow-scaffolding.md"
  ],
  "architectureNotes": "architecture-notes.md",
  "patternReferences": [
    {
      "story": "US-003",
      "pattern": "Real-time updates",
      "reference": "rugby-crm/src/features/notifications",
      "reason": "Similar WebSocket + polling pattern"
    }
  ]
}

Key Guidelines

1. Pattern Discovery:

• Always check scope.references for existing implementations
• Reference file paths explicitly (helps execution phase locate code)
• Call out shared utilities and middleware

2. File Organization:

• Follow existing project structure
• Use established naming conventions (services/, controllers/, components/)
• Group related files by feature

3. Import Statements:

• Include enough detail to guide code generation
• Use actual import paths from codebase
• Call out shared packages or internal utilities

4. Database Modeling:

• Include ORM-specific syntax if applicable (Prisma, TypeORM, etc.)
• Specify indexes for performance-critical queries
• Document relationships between models

5. Conciseness:

• Total architect work: 50-100 lines max
• Point to reference repos instead of duplicating patterns
• Use tables/lists, not prose

Integration

Called by:

• /build command (after design-phase approval)
• /ms command (complexity ≥15, after design-phase approval)

Calls:

• architect agent (enrichment mode, complexity-aware model)

Next step: execution-phase (workers read enriched stories and implement)

Example Workflow

Design Phase Output:
  stories/
    US-001-database-schema.md
    US-002-contracts-types.md
    US-003-workflow-scaffolding.md

Architect Phase (this skill):
  → Read scope: targets=[metasaver-com], references=[rugby-crm]
  → For each story:
    - Check rugby-crm for similar patterns
    - Add API endpoints
    - List files to create/modify
    - Add import statements
    - Specify database models
    - Reference components
  → Save enriched stories
  → Create architecture-notes.md for cross-cutting concerns

Output:
  {
    enrichedStories: [US-001-enriched.md, US-002-enriched.md, ...],
    architectureNotes: "architecture-notes.md",
    patternReferences: [...]
  }

Execution Phase:
  → Workers read enriched stories
  → Execute story-by-story with full architecture context
  → Create/modify files listed in Architecture Details

Common Patterns to Enrich

Checklist for Architect

• All stories have Architecture Details section
• API endpoints match project conventions
• Files paths use correct directory structure
• Import statements are accurate and complete
• Database models include required fields and indexes
• Component names follow naming conventions
• Pattern references point to actual code locations
• Cross-cutting concerns documented (if complex)
• No duplication of patterns (reference instead)