Skill: architecture

What I do

I enforce clean architecture: layer separation (domain → service → repository → handler), dependency direction (inward only), and boundary rules that keep the codebase maintainable as it grows.

When to use me

• Designing new packages, intents, or modules
• Reviewing code for layer boundary violations
• Deciding where new logic belongs (domain vs service vs handler)
• Structuring Go projects with clean dependency flow
• Diagnosing tight coupling or circular dependencies

Core principles

1. Dependencies point inward — Domain knows nothing about HTTP, databases, or frameworks
2. Layer isolation — Each layer has a single responsibility; no layer skipping
3. Interface boundaries — Layers communicate through interfaces defined by the consumer
4. Domain is king — Business rules live in domain; everything else is infrastructure
5. Package by feature — Group by capability (user/, order/), not by type (models/, handlers/)

Patterns & examples

Layer responsibilities:

Dependency flow in Go:

// domain/ — no imports from other layers
type User struct {
    ID    string
    Email string
    Name  string
}

type UserRepository interface {
    Save(ctx context.Context, user *User) error
    FindByEmail(ctx context.Context, email string) (*User, error)
}

// service/ — depends only on domain
type UserService struct {
    repo domain.UserRepository  // interface, not concrete
}

func (s *UserService) Register(ctx context.Context, email, name string) error {
    user := &domain.User{Email: email, Name: name}
    return s.repo.Save(ctx, user)
}

// handler/ — depends on service
func (h *Handler) RegisterUser(w http.ResponseWriter, r *http.Request) {
    // Decode request, call service, encode response
    err := h.svc.Register(r.Context(), req.Email, req.Name)
}

// infrastructure/ — implements domain interfaces
type GORMUserRepo struct{ db *gorm.DB }
func (r *GORMUserRepo) Save(ctx context.Context, u *domain.User) error { ... }

Package structure (feature-based):

intent/
├── user/
│   ├── domain/       # entities, value objects, interfaces
│   ├── service/      # use cases
│   ├── repository/   # data access implementation
│   └── handler/      # HTTP handlers
├── order/
│   ├── domain/
│   ├── service/
│   └── ...

Boundary validation checklist:

• Domain imports: only stdlib (fmt, errors, time)
• Service imports: domain only
• Handler imports: service only (never domain directly for persistence)
• Repository imports: domain (for interfaces/entities) + infrastructure (GORM, etc.)

Anti-patterns to avoid

• ❌ Handler calling repository directly — Skips business logic; service layer exists for a reason
• ❌ Domain importing infrastructure — Domain must not know about GORM, HTTP, or external services
• ❌ Circular dependencies — Package A imports B, B imports A; restructure with interfaces
• ❌ God package — Single models/ package with everything; package by feature instead
• ❌ Leaking implementation — Returning GORM models from service layer; map to domain types

KB Reference

~/vaults/baphled/3. Resources/Knowledge Base/AI Development System/Skills/Domain-Architecture/Architecture.md

Related skills

• domain-modeling - Designing entities and value objects in the domain layer
• service-layer - Orchestrating use cases in the service layer
• design-patterns - Patterns that support architectural boundaries
• clean-code - Code quality within each layer
• modular-design - Unit-level composability and testability within each layer