Skip to main content
ClaudeWave
EN, English
Skill1.4k repo starsupdated 22d ago

designing-architecture

# Claude Code: Designing Architecture This skill guides developers through systematic software architecture design by providing decision frameworks, pattern selection tables, and workflow checklists. Use it when planning system designs, choosing between monolithic and microservices approaches, structuring new projects, making technical decisions about modules and components, or selecting appropriate architectural patterns based on project size and team capabilities.

Install in Claude Code
Copy
git clone --depth 1 https://github.com/CloudAI-X/claude-workflow-v2 /tmp/designing-architecture && cp -r /tmp/designing-architecture/skills/designing-architecture ~/.claude/skills/designing-architecture
Then start a new Claude Code session; the skill loads automatically.

SKILL.md

# Designing Architecture

### When to Load

- **Trigger**: System design, module structure, new project scaffolding, choosing architecture patterns
- **Skip**: Simple bug fixes or minor code changes that don't affect architecture

## Architecture Decision Workflow

Copy this checklist and track progress:

```
Architecture Design Progress:
- [ ] Step 1: Understand requirements and constraints
- [ ] Step 2: Assess project size and team capabilities
- [ ] Step 3: Select architecture pattern
- [ ] Step 4: Define directory structure
- [ ] Step 5: Document trade-offs and decision
- [ ] Step 6: Validate against decision framework
```

## Pattern Selection Guide

### By Project Size

| Size              | Recommended Pattern               |
| ----------------- | --------------------------------- |
| Small (<10K LOC)  | Simple MVC/Layered                |
| Medium (10K-100K) | Clean Architecture                |
| Large (>100K)     | Modular Monolith or Microservices |

### By Team Size

| Team      | Recommended                  |
| --------- | ---------------------------- |
| 1-3 devs  | Monolith with clear modules  |
| 4-10 devs | Modular Monolith             |
| 10+ devs  | Microservices (if justified) |

## Common Patterns

### 1. Layered Architecture

```
┌─────────────────────────────┐
│       Presentation          │  ← UI, API Controllers
├─────────────────────────────┤
│       Application           │  ← Use Cases, Services
├─────────────────────────────┤
│         Domain              │  ← Business Logic, Entities
├─────────────────────────────┤
│      Infrastructure         │  ← Database, External APIs
└─────────────────────────────┘
```

**Use when**: Simple CRUD apps, small teams, quick prototypes

### 2. Clean Architecture

```
┌─────────────────────────────────────┐
│            Frameworks & Drivers      │
│  ┌─────────────────────────────┐    │
│  │     Interface Adapters       │    │
│  │  ┌─────────────────────┐    │    │
│  │  │   Application       │    │    │
│  │  │  ┌─────────────┐    │    │    │
│  │  │  │   Domain    │    │    │    │
│  │  │  └─────────────┘    │    │    │
│  │  └─────────────────────┘    │    │
│  └─────────────────────────────┘    │
└─────────────────────────────────────┘
```

**Use when**: Complex business logic, long-lived projects, testability is key

### 3. Hexagonal (Ports & Adapters)

```
        ┌──────────┐
        │ HTTP API │
        └────┬─────┘
             │ Port
    ┌────────▼────────┐
    │                 │
    │   Application   │
    │     Core        │
    │                 │
    └────────┬────────┘
             │ Port
        ┌────▼─────┐
        │ Database │
        └──────────┘
```

**Use when**: Need to swap external dependencies, multiple entry points

### 4. Event-Driven Architecture

```
Producer → Event Bus → Consumer
              │
              ├─→ Consumer
              │
              └─→ Consumer
```

**Use when**: Loose coupling needed, async processing, scalability

### 5. CQRS (Command Query Responsibility Segregation)

```
┌─────────────┐      ┌─────────────┐
│  Commands   │      │   Queries   │
│  (Write)    │      │   (Read)    │
└──────┬──────┘      └──────┬──────┘
       │                    │
       ▼                    ▼
  Write Model          Read Model
       │                    │
       └────────┬───────────┘
                ▼
           Event Store
```

**Use when**: Different read/write scaling, complex domains, event sourcing

## Directory Structure Patterns

### Feature-Based (Recommended for medium+)

```
src/
├── features/
│   ├── users/
│   │   ├── api/
│   │   ├── components/
│   │   ├── hooks/
│   │   ├── services/
│   │   └── types/
│   └── orders/
│       ├── api/
│       ├── components/
│       └── ...
├── shared/
│   ├── components/
│   ├── hooks/
│   └── utils/
└── app/
    └── ...
```

### Layer-Based (Simple apps)

```
src/
├── controllers/
├── services/
├── models/
├── repositories/
└── utils/
```

## Decision Framework

When making architectural decisions, evaluate against these criteria:

1. **Simplicity** - Start simple, evolve when needed
2. **Team Skills** - Match architecture to team capabilities
3. **Requirements** - Let business needs drive decisions
4. **Scalability** - Consider growth trajectory
5. **Maintainability** - Optimize for change

## Trade-off Analysis Template

Use this template to document architectural decisions:

```markdown
## Decision: [What we're deciding]

### Context

[Why this decision is needed now]

### Options Considered

1. Option A: [Description]
2. Option B: [Description]

### Trade-offs

| Criteria         | Option A | Option B |
| ---------------- | -------- | -------- |
| Complexity       | Low      | High     |
| Scalability      | Medium   | High     |
| Team familiarity | High     | Low      |

### Decision

We chose [Option] because [reasoning].

### Consequences

- [What this enables]
- [What this constrains]
```

## Validation Checklist

After selecting an architecture, validate against:

```
Architecture Validation:
- [ ] Matches project size and complexity
- [ ] Aligns with team skills and experience
- [ ] Supports current requirements
- [ ] Allows for anticipated growth
- [ ] Dependencies flow inward (core has no external deps)
- [ ] Clear boundaries between modules/layers
- [ ] Testing strategy is feasible
- [ ] Trade-offs are documented
```

If validation fails, reconsider the pattern selection or adjust the implementation approach.