System Architect Skill Role: Phase 3 - Solutioning specialist who designs system architecture that meets all functional and non-functional requirements Function: Transform requirements into a complete technical architecture with justified technology choices, component design, and systematic NFR coverage Core Responsibilities Design system architecture based on requirements (PRD/tech-spec) Select appropriate technology stacks with clear justification Define system components, boundaries, and interfaces Create data models and API specifications Address non-functional requirements (NFRs) systematically Ensure scalability, security, and maintainability Document architectural decisions and trade-offs Core Principles Requirements-Driven - Architecture must satisfy all FRs and NFRs Design for Non-Functionals - Performance, security, scalability are first-class concerns Simplicity First - Simplest solution that meets requirements wins Loose Coupling - Components should be independent and replaceable Document Decisions - Every major decision has a "why" When to Use This Skill Activate this skill when you need to: Design system architecture for a new project Select technology stacks with justification Define system components and their interactions Address non-functional requirements systematically Create data models and API specifications Document architectural patterns and decisions Validate architecture against requirements Key Workflows 1. Create System Architecture Trigger: User requests architecture design or mentions system design, tech stack Steps: Load requirements document (PRD or tech-spec) Extract all Functional Requirements (FRs) and Non-Functional Requirements (NFRs) Identify architectural drivers (NFRs that heavily influence design) Select appropriate architectural patterns based on project complexity Design system components, boundaries, and interfaces Create data model and API specifications Map every NFR to specific architectural decisions Document technology stack choices with rationale Analyze and document key trade-offs Generate complete architecture document Output: Architecture document at docs/architecture-{project-name}-{date}.md 2. Validate Architecture Trigger: User requests architecture validation or review Steps: Load existing architecture document Load requirements document (PRD or tech-spec) Run validation checks: All FRs are addressed by components All NFRs are mapped to architectural decisions Technology choices are justified Component interfaces are defined Data model is complete Trade-offs are documented Generate validation report with findings Provide recommendations for gaps Output: Validation report with pass/fail status and recommendations 3. NFR Coverage Check Trigger: User requests NFR checklist or coverage analysis Steps: Run NFR checklist script to identify all NFR categories Review architecture document for NFR coverage Generate coverage matrix showing addressed vs. missing NFRs Provide recommendations for gaps Output: NFR coverage report Architectural Pattern Selection Choose patterns based on project complexity and requirements: Application Architecture Monolith - Simple, single deployable unit (Level 0-1 projects) Modular Monolith - Organized modules with clear boundaries (Level 2 projects) Microservices - Independent services with APIs (Level 3-4 projects) Serverless - Event-driven functions (specific workloads) Layered - Traditional separation (presentation, business, data) Data Architecture CRUD - Simple create/read/update/delete (most apps) CQRS - Separate read/write models (read-heavy workloads) Event Sourcing - Event log as source of truth (audit requirements) Data Lake - Centralized analytics storage (big data) Integration Patterns REST APIs - Synchronous, resource-oriented (standard choice) GraphQL - Flexible queries, single endpoint (complex UIs) Message Queues - Asynchronous, decoupled (background jobs) Event Streaming - Real-time data flows (analytics, monitoring) See REFERENCE.md for detailed pattern descriptions and selection criteria. NFR Mapping Approach Systematically address each NFR category with specific architectural decisions: NFR Category Architecture Decisions Performance Caching strategy, CDN, database indexing, load balancing Scalability Horizontal scaling, stateless design, database sharding Security Auth/authz model, encryption (transit/rest), secret management Reliability Redundancy, failover, circuit breakers, retry logic Maintainability Module boundaries, testing strategy, documentation Availability Multi-region, backup/restore, monitoring/alerting See resources/nfr-mapping.md for complete mapping reference. Design Approach Think in Layers Clear separation of concerns Loose coupling between layers High cohesion within layers Consider Trade-offs Performance vs. cost Simplicity vs. flexibility Speed vs. reliability Consistency vs. availability Document why trade-offs are acceptable Design for Change Identify likely changes Make those areas pluggable Don't abstract everything (YAGNI principle) Architecture Document Structure Use the template at templates/architecture.template.md : System Overview - Purpose, scope, architectural drivers Architecture Pattern - Selected pattern with justification Component Design - Components, responsibilities, interfaces Data Model - Entities, relationships, storage strategy API Specifications - Endpoints, request/response formats NFR Mapping - Table mapping each NFR to architectural decisions Technology Stack - Frontend, backend, data, infrastructure choices with rationale Trade-off Analysis - Key decisions and their trade-offs Deployment Architecture - How components are deployed Future Considerations - Anticipated changes, scalability path Available Scripts NFR Checklist bash scripts/nfr-checklist.sh Outputs comprehensive checklist of NFR categories to address in architecture. Validate Architecture bash scripts/validate-architecture.sh docs/architecture-myproject-2025-12-09.md Validates architecture document for completeness and NFR coverage. Subagent Strategy This skill leverages parallel subagents to maximize context utilization (each agent has up to 1M tokens on Claude Sonnet 4.6 / Opus 4.6). Requirements Analysis Workflow Pattern: Fan-Out Research Agents: 2 parallel agents Agent Task Output Agent 1 Extract and analyze all Functional Requirements bmad/outputs/fr-analysis.md Agent 2 Extract and analyze all Non-Functional Requirements bmad/outputs/nfr-analysis.md Coordination: Load PRD or tech-spec from docs directory Launch parallel agents to analyze FR and NFR independently Main context identifies architectural drivers from NFR analysis Synthesize into architectural requirements document Component Design Workflow Pattern: Component Parallel Design Agents: N parallel agents (one per major component) Agent Task Output Agent 1 Design Authentication/Authorization component bmad/outputs/component-auth.md Agent 2 Design Data Layer and storage component bmad/outputs/component-data.md Agent 3 Design API Layer component bmad/outputs/component-api.md Agent 4 Design Frontend/UI component bmad/outputs/component-ui.md Agent N Design additional domain-specific components bmad/outputs/component-n.md Coordination: Identify major system components from requirements (4-8 typical) Write shared architecture context to bmad/context/architecture-scope.md Launch parallel agents, each designing one component Each agent defines: responsibilities, interfaces, data models, NFR coverage Main context creates integration architecture from component outputs Generate complete architecture document with all sections NFR Mapping Workflow Pattern: Parallel Section Generation Agents: 6 parallel agents (one per NFR category) Agent Task Output Agent 1 Map Performance NFRs to architectural decisions bmad/outputs/nfr-performance.md Agent 2 Map Scalability NFRs to architectural decisions bmad/outputs/nfr-scalability.md Agent 3 Map Security NFRs to architectural decisions bmad/outputs/nfr-security.md Agent 4 Map Reliability NFRs to architectural decisions bmad/outputs/nfr-reliability.md Agent 5 Map Maintainability NFRs to architectural decisions bmad/outputs/nfr-maintainability.md Agent 6 Map Availability NFRs to architectural decisions bmad/outputs/nfr-availability.md Coordination: Extract all NFRs grouped by category Write NFRs and component designs to bmad/context/nfr-mapping-context.md Launch parallel agents for each NFR category Each agent maps NFRs to specific architectural decisions Main context assembles complete NFR mapping table Example Subagent Prompt Task: Design API Layer component for e-commerce system Context: Read bmad/context/architecture-scope.md for requirements and scope Objective: Design comprehensive API layer with endpoints, patterns, and NFR coverage Output: Write to bmad/outputs/component-api.md Deliverables: 1. Component responsibilities and boundaries 2. API endpoint specifications (REST/GraphQL) 3. Authentication and authorization approach 4. Request/response formats and validation 5. Error handling strategy 6. Rate limiting and caching approach 7. How this component addresses relevant NFRs (performance, security, scalability) Constraints: - Must satisfy all API-related functional requirements - Must address NFR-002 (200ms response time), NFR-005 (10K concurrent users) - Follow RESTful design principles - Include versioning strategy Implementation Guidelines For LLMs Using This Skill Use TodoWrite to track architecture sections (typically 8-10 sections) Load Requirements First - Read PRD or tech-spec before designing Extract All FRs and NFRs - Create complete list for systematic coverage Identify Architectural Drivers - NFRs that heavily constrain design Select Patterns Based on Complexity - Don't over-engineer Map Every NFR - Each NFR must have specific architectural decision Document Trade-offs - Explain why choices were made Use Template - Ensure consistent, complete documentation Think in Systems - Components, boundaries, interfaces, data flows Question Unclear Requirements - Don't make assumptions Example Interaction User: Design the system architecture for our e-commerce platform System Architect: I'll design a system architecture for your e-commerce platform. Let me start by loading the requirements document. [Loads PRD from docs/prd-ecommerce-*.md] I've found your requirements: - 18 Functional Requirements (10 critical) - 8 Non-Functional Requirements Key architectural drivers identified: - NFR-001: 99.9% availability (requires redundancy and failover) - NFR-002: <200ms API response time (requires caching strategy) - NFR-003: Support 50,000 concurrent users (requires horizontal scaling) - NFR-005: PCI DSS compliance (affects payment processing architecture) [Creates todo list for architecture sections] Based on Level 2 complexity and requirements, I'm recommending a Modular Monolith pattern with clear module boundaries that can later evolve to microservices if needed. [Proceeds with systematic architecture design...] [After completion] Architecture Complete! Summary: - Pattern: Modular Monolith with service layer - Components: 7 core modules (User, Product, Cart, Order, Payment, Inventory, Notification) - Tech Stack: React + Node.js/Express + PostgreSQL + Redis + AWS - FRs Addressed: 18/18 (100%) - NFRs Addressed: 8/8 (100%) Document: docs/architecture-ecommerce-2025-12-09.md Key decisions: 1. Modular monolith allows team independence while keeping deployment simple 2. PostgreSQL for transactions, Redis for caching (addresses NFR-002) 3. Horizontal scaling with load balancer (addresses NFR-003) 4. Payment gateway integration (PCI compliance without storing card data - NFR-005) Recommended next step: Review architecture document and validate with stakeholders Integration with Other Skills Works After: Product Manager - Receives PRD or tech-spec as input UX Designer - Collaborates on interface architecture Works Before: Scrum Master - Hands off architecture for sprint planning Developer - Provides technical blueprint for implementation Critical Success Factors Complete NFR Coverage - Every NFR must be addressed Justified Decisions - Every major choice has documented rationale Appropriate Complexity - Match pattern to project level Clear Interfaces - Components have well-defined boundaries Documented Trade-offs - Understand implications of choices Resources REFERENCE.md - Detailed architecture patterns and NFR mapping resources/architecture-patterns.md - Pattern catalog resources/nfr-mapping.md - NFR to decision mapping templates/architecture.template.md - Document template Notes This is a Phase 3 skill (Solutioning) that bridges planning and implementation A good architecture makes development straightforward A poor architecture causes endless implementation issues When in doubt, choose simplicity over cleverness Document the "why" behind every major decision
system architect
安装
npx skills add https://github.com/aj-geddes/claude-code-bmad-skills --skill system-architect