Umbrella Multi-Repo Architecture Detector
When This Skill Activates
Activates when user describes:
- •Multiple repos: "3 repos", "frontend repo", "backend repo", "shared library"
- •Architecture patterns: "monorepo with services", "microservices", "multi-repo"
- •Explicit splits: "FE/BE/Shared", "frontend/backend/common"
- •GitHub URLs for multiple repositories
My Role
When I detect a multi-repo architecture in the user's prompt:
- •Acknowledge the architecture with detected repos
- •Explain project-scoped user stories (US-FE-, US-BE-, US-SHARED-*)
- •Guide the init flow for proper setup
- •Route to PM agent with multi-repo context
Detection Patterns
| Pattern | Example | Detected As |
|---|---|---|
| Repo count | "3 repos", "multiple repos" | Multi-repo intent |
| Frontend repo | "Frontend repo", "UI repo", "web app" | Type: frontend, Prefix: FE |
| Backend repo | "Backend API repo", "server", "API" | Type: backend, Prefix: BE |
| Shared repo | "Shared library", "common types" | Type: shared, Prefix: SHARED |
| Mobile repo | "Mobile app", "iOS/Android" | Type: mobile, Prefix: MOBILE |
| Infra repo | "Infrastructure", "Terraform" | Type: infrastructure, Prefix: INFRA |
Project-Scoped User Stories
When user describes multi-repo, user stories MUST be prefixed:
## Instead of (generic): US-001: User Registration US-002: Registration API US-003: Validation Schema ## Generate (project-scoped): US-FE-001: User Registration Form - Related repo: frontend - Keywords: form, UI, validation display US-BE-001: Registration API Endpoint - Related repo: backend - Keywords: API, endpoint, database US-SHARED-001: Registration Validation Schema - Related repo: shared - Keywords: validator, schema, types
Cross-Cutting User Stories
For features that span multiple repos, use cross-project tagging:
US-AUTH-001: OAuth Integration
- Tags: ["cross-project", "frontend", "backend"]
- Creates linked issues in: FE repo, BE repo
- Child stories:
- US-FE-002: OAuth Login Button (frontend)
- US-BE-002: OAuth Token Validation (backend)
Setup Flow Guidance
When multi-repo detected, guide user through options:
Option 1: Clone from GitHub (Recommended)
You have existing repos? Let's clone them: 1. Provide GitHub URLs (comma-separated or one per line) 2. Each repo gets its own .specweave/ configuration 3. Each repo syncs to its own GitHub issues
Option 2: Create New Repos
Creating fresh repos: 1. I'll create repos on GitHub for you 2. Each gets initialized with .specweave/ 3. External tool sync configured per repo
Option 3: Initialize Local Folders
Have local folders already? 1. Point me to each folder 2. I'll initialize .specweave/ in each 3. Configure external tools per repo
Umbrella Repo Structure
umbrella-project/ # Optional parent repo
├── .specweave/
│ ├── config.json # umbrella config with childRepos[]
│ └── docs/ # High-level PRD, roadmap only
│
├── my-app-fe/ # Frontend repo (cloned/created)
│ └── .specweave/
│ ├── config.json # sync → my-app-fe GitHub issues
│ └── increments/
│ └── 0001-feature/
│ └── spec.md # Only US-FE-* stories
│
├── my-app-be/ # Backend repo (cloned/created)
│ └── .specweave/
│ └── ... # sync → my-app-be GitHub issues
│
└── my-app-shared/ # Shared repo (cloned/created)
└── .specweave/
└── ... # sync → my-app-shared GitHub issues
Project ID Strategy
CRITICAL: The id field MUST match your canonical source name - no arbitrary abbreviations!
| Scenario | ID Source | Example |
|---|---|---|
| 1:1 Repo Mapping | Exact repo name | sw-qr-menu-fe |
| JIRA Project | Project key (lowercase) | WEBAPP → webapp |
| ADO Project | Project name (kebab-case) | Frontend Team → frontend-team |
| Area Path | Last segment (kebab-case) | Product\Web → web |
✅ CORRECT: id matches repo name id: "sw-qr-menu-fe" path: "./sw-qr-menu-fe" githubUrl: "https://github.com/user/sw-qr-menu-fe" ❌ WRONG: arbitrary abbreviation id: "fe" ← What if you have 2 frontend repos? path: "./sw-qr-menu-fe"
Note: The prefix (for user stories like US-FE-001) can be short even if id is long:
- •
id: "sw-qr-menu-fe"(full repo name) - •
prefix: "FE"(short, for user story IDs)
Config Example
Parent umbrella config (.specweave/config.json):
{
"umbrella": {
"enabled": true,
"childRepos": [
{
"id": "sw-qr-menu-fe",
"path": "./sw-qr-menu-fe",
"prefix": "FE",
"githubUrl": "https://github.com/myorg/sw-qr-menu-fe"
},
{
"id": "sw-qr-menu-be",
"path": "./sw-qr-menu-be",
"prefix": "BE",
"githubUrl": "https://github.com/myorg/sw-qr-menu-be"
},
{
"id": "sw-qr-menu-shared",
"path": "./sw-qr-menu-shared",
"prefix": "SHARED",
"githubUrl": "https://github.com/myorg/sw-qr-menu-shared"
}
]
}
}
JIRA-based project (when JIRA is source of truth):
{
"umbrella": {
"enabled": true,
"childRepos": [
{
"id": "webapp",
"path": "./frontend",
"prefix": "WEBAPP",
"jiraProject": "WEBAPP",
"githubUrl": "https://github.com/myorg/frontend"
}
]
}
}
Child repo config (sw-qr-menu-fe/.specweave/config.json):
{
"project": {
"name": "QR Menu Frontend",
"prefix": "FE"
},
"sync": {
"activeProfile": "github",
"profiles": {
"github": {
"provider": "github",
"config": {
"owner": "myorg",
"repo": "sw-qr-menu-fe"
}
}
}
}
}
Response Template
When I detect multi-repo intent, respond:
I detected a **multi-repo architecture** in your description: **Detected Repos:** - Frontend (prefix: FE) - [matched keywords] - Backend (prefix: BE) - [matched keywords] - Shared (prefix: SHARED) - [matched keywords] **User Story Format:** User stories will be project-scoped: - `US-FE-001`: Frontend stories - `US-BE-001`: Backend stories - `US-SHARED-001`: Shared library stories **Setup Options:** 1. **Clone from GitHub** - Provide URLs, I'll clone and initialize each 2. **Create new repos** - I'll create on GitHub and initialize 3. **Initialize local folders** - Point to existing folders Which would you like to do?
Keywords for Story Routing
| Keywords | Routes To | Prefix |
|---|---|---|
| UI, component, page, form, view, theme, drag-drop, builder | Frontend | FE |
| API, endpoint, CRUD, webhook, notification, analytics | Backend | BE |
| schema, validator, types, utilities, localization | Shared | SHARED |
| iOS, Android, mobile, push notification | Mobile | MOBILE |
| Terraform, K8s, Docker, CI/CD | Infrastructure | INFRA |
Saving Changes Across Repos
Use /sw:save to commit and push changes across all repos at once:
# Save all repos with same commit message /sw:save "feat: Add user authentication" # Preview what would happen /sw:save --dry-run # Save specific repos only /sw:save "fix: Bug fixes" --repos frontend,backend
Features:
- •Auto-discovers nested repos - Scans
repositories/,packages/,services/,apps/,libs/for.gitdirectories (up to 4 levels deep) - •Auto-detects repos with changes
- •Sets up remotes if missing (prompts for URL or uses umbrella config)
- •Commits with same message to all repos
- •Pushes to origin
- •Skips repos with no changes
Auto-Discovery (No Config Required):
Even without umbrella config, /sw:save automatically finds all nested repos:
my-project/ ├── repositories/ │ ├── frontend/.git # ← Auto-discovered │ ├── backend/.git # ← Auto-discovered │ └── shared/.git # ← Auto-discovered └── .git # ← Parent repo included
All 4 repos will be committed and pushed with a single /sw:save command!
Important Notes
- •Each repo is independent - Own
.specweave/, own increments, own external tool sync - •Parent repo is optional - Can have umbrella config or just independent repos
- •User stories MUST have project prefix - Never generate generic
US-001in multi-repo mode - •Cross-project stories get special handling - Tagged and linked across repos
- •Use
/sw:save- Single command to save changes across all repos
Project-Specific Learnings
Before starting work, check for project-specific learnings:
# Check if skill memory exists for this skill cat .specweave/skill-memories/umbrella-repo-detector.md 2>/dev/null || echo "No project learnings yet"
Project learnings are automatically captured by the reflection system when corrections or patterns are identified during development. These learnings help you understand project-specific conventions and past decisions.