---

name: clean-architecture description: Clean architecture principles for separating concerns. Use when designing new modules, creating interfaces, defining system boundaries, refactoring mixed concerns, or reviewing code for architecture violations. Applies to all languages. user-invocable: false

<!-- This skill follows the Agent Skills open standard: https://agentskills.io -->

Clean Architecture

When working on architecture decisions, module design, or separation of concerns, read and follow the policy guide:

Policy: docs/policies/clean-architecture.md

When This Applies

• Designing new modules, services, or bounded contexts
• Creating interface contracts between layers
• Choosing between repository, service, gateway, or strategy patterns
• Organizing code into module-based folder structures
• Reviewing code for architecture layer violations
• Deciding where to place new functionality

Key Rules (from CLAUDE.md)

Dependencies point inward. Inner layers MUST NOT depend on outer layers:

Business Logic (inner)  -- Pure functions, domain logic, validation
  depends on
Interface (boundary)    -- Contracts between layers (I{Name} interfaces)
  implemented by
System Implementation (outer) -- External I/O (*.system.ts files)

Quick Reference

• One interface per system boundary (HTTP client, database, file system)
• Module folders when 5+ related files share a prefix
• Barrel exports via index.ts for public API
• Name by business capability, not implementation tool (e.g., ISecretStore not IOnePasswordClient)

Read the full guide for patterns, examples, and folder organization.