AgentSkillsCN

Architecture

架构设计

SKILL.md
--- frontmatter
name: "Architecture"
id: "architecture"
version: "1.0"
claude_subagent: "backend-architect"
fallback: "self-check"
min_tad_version: "2.1"
platforms: ["claude", "codex", "gemini"]

Architecture Skill

Purpose

Review system architecture, design patterns, data flow, and structural decisions for scalability, maintainability, and best practices.

When to Use

  • •During Gate 2 (design completeness)
  • •For new feature architecture
  • •For refactoring decisions
  • •For microservices/module boundaries
  • •For database schema design

Checklist

Critical (P0) - Must Pass

  • • Clear separation of concerns
  • • No circular dependencies
  • • Data flow is traceable
  • • Error boundaries defined
  • • No architectural anti-patterns

Important (P1) - Should Pass

  • • Consistent design patterns used
  • • Module boundaries well-defined
  • • Dependencies injected (not hardcoded)
  • • Configuration externalized
  • • Scalability considered

Nice-to-have (P2) - Informational

  • • Documentation reflects architecture
  • • ADRs (Architecture Decision Records) present
  • • Monitoring hooks planned
  • • Feature flags for gradual rollout
  • • Migration path documented

Suggestions (P3) - Optional

  • • Future scaling considerations
  • • Technology upgrade paths
  • • Alternative approaches considered
  • • Technical debt items identified

Pass Criteria

LevelRequirement
P0All items pass
P1Max 2 failures
P2Informational
P3Optional

Evidence Output

Path: .tad/evidence/reviews/{date}-architecture-{task}.md

Execution Contract

  • •Input: file_paths[], diagrams[], context{}
  • •Output: {passed: bool, findings: [{severity, category, component, description, recommendation}], evidence_path: string}
  • •Timeout: 240s
  • •Parallelizable: true

Claude Enhancement

When running on Claude Code, call subagent backend-architect for deeper analysis. Reference: .tad/templates/output-formats/architecture-review-format.md

Architecture Review Categories

Structural Patterns

  • •Layered architecture
  • •Clean architecture
  • •Hexagonal/ports-adapters
  • •Event-driven
  • •Microservices boundaries

Data Architecture

  • •Database schema design
  • •Data flow patterns
  • •Caching strategies
  • •Data consistency models
  • •Migration strategies

Integration Patterns

  • •API design
  • •Message queues
  • •Event sourcing
  • •CQRS considerations
  • •Third-party integrations

Quality Attributes

  • •Scalability
  • •Maintainability
  • •Testability
  • •Observability
  • •Reliability

Anti-Patterns to Avoid

  • •God classes/modules
  • •Circular dependencies
  • •Tight coupling
  • •Leaky abstractions
  • •Premature optimization