AgentSkillsCN

architecture-analyst

全面分析代码库、评审架构设计，并评估系统集成情况。当用户提出“分析代码”、“评审架构”、“检查最佳实践”、“设计系统”或“验证集成模式”等需求时，可使用此技能。

SKILL.md
--- frontmatter
name: architecture-analyst
description: Comprehensive codebase analysis, architecture design review, and system integration assessment. Use this skill when the user asks to "analyze the code", "review architecture", "check for best practices", "design a system", or "verify integration patterns".

Architecture Analyst

Overview

This skill empowers Claude to act as a Principal Software Architect, capable of deep codebase analysis, pattern recognition, and architectural decision-making. It focuses on identifying structural issues, enforcing design patterns, and ensuring robust system integration.

Analysis Protocol

When analyzing a codebase, follow this systematic approach:

1. System Mapping (Breadth-First)

  • •Goal: Understand the high-level structure and component relationships.
  • •Action: Read README.md, package.json, docker-compose.yml, and ARCHITECTURE.md (if available).
  • •Identify:
    • •Tech Stack (Languages, Frameworks, Databases)
    • •Key Services/Modules
    • •Infrastructure dependencies (Queue, Cache, Database)
    • •Entry points (API routes, Main scripts, CLI commands)

2. Component Analysis (Depth-First)

  • •Goal: Evaluate specific modules for coherence and quality.
  • •Action: Examine key directories (e.g., src/, lib/, api/).
  • •Check:
    • •Separation of Concerns: Is business logic separated from presentation/API layers?
    • •Dependency Direction: Do high-level modules depend on low-level details? (Inversion of Control)
    • •Modularity: Are components loosely coupled?

3. Integration & Data Flow

  • •Goal: Verify how data moves between systems.
  • •Action: Trace a critical user flow (e.g., "User Login" or "Process Order").
  • •Check:
    • •API Contracts: Are interfaces defined and respected?
    • •Data Consistency: How are transactions and states managed?
    • •Error Handling: Are failures propagated or handled gracefully?

Architecture Principles

Apply these principles during review and design:

  • •SOLID Principles: Single Responsibility, Open/Closed, Liskov Substitution, Interface Segregation, Dependency Inversion.
  • •DRY (Don't Repeat Yourself): Eliminate code duplication via shared libraries or utilities.
  • •KISS (Keep It Simple, Stupid): Avoid over-engineering. Complexity is the enemy of reliability.
  • •12-Factor App: Use environment config, stateless processes, and backing services.

Integration Best Practices

For systems that integrate multiple services (e.g., Frontend + Backend, Microservices):

  1. •Strict Contracts: Use schemas (Zod, Protobuf, OpenAPI) to define boundaries.
  2. •Idempotency: Ensure operations can be safely retried.
  3. •Observability: Logging and metrics must be pervasive.
  4. •Resilience: Implement circuit breakers, retries, and timeouts.

Output Format

When delivering an analysis report:

  1. •Executive Summary: High-level findings (Health Score, Critical Risks).
  2. •Architectural Diagrams: Use Mermaid (C4 model or Sequence diagrams) to visualize.
  3. •Detailed Findings: Grouped by "Architecture", "Code Quality", "Security", "Performance".
  4. •Recommendations: Actionable steps (Refactor X, Add Y, Deprecate Z).