AgentSkillsCN

architecture

架构决策框架。包括需求分析、权衡取舍评估，以及 ADR 文档的编写。当你需要做出架构决策，或对系统设计进行深入分析时，可选用此技能。

SKILL.md
--- frontmatter
name: architecture
description: Architectural decision-making framework. Requirements analysis, trade-off evaluation, ADR documentation. Use when making architecture decisions or analyzing system design.
allowed-tools: Read, Glob, Grep

Architecture Decision Framework

"Requirements drive architecture. Trade-offs inform decisions. ADRs capture rationale."

🎯 Selective Reading Rule

Read ONLY files relevant to the request! Check the content map, find what you need.

FileDescriptionWhen to Read
context-discovery.mdQuestions to ask, project classificationStarting architecture design
trade-off-analysis.mdADR templates, trade-off frameworkDocumenting decisions
pattern-selection.mdDecision trees, anti-patternsChoosing patterns
examples.mdMVP, SaaS, Enterprise examplesReference implementations
patterns-reference.mdQuick lookup for patternsPattern comparison

🔗 Related Skills

SkillUse For
@[skills/database-design]Database schema design
@[skills/api-patterns]API design patterns
@[skills/deployment-procedures]Deployment architecture

Core Principle

"Simplicity is the ultimate sophistication."

  • •Start simple
  • •Add complexity ONLY when proven necessary
  • •You can always add patterns later
  • •Removing complexity is MUCH harder than adding it

📐 System Design Template (Staff Level)

Don't start with boxes. Start with math and requirements.

  1. •Requirements & Constraints
    • •Functional: "User clips video", "System generates subtitles"
    • •Non-Functional: "Wait time < 20s", "99.9% Availability", "Budget < $500/mo"
  2. •Back-of-Envelope Math (Capacity)
    • •Formula: $QPS = DailyActiveUsers \times ActionsPerUser / 86400$
    • •Formula: $Storage = WritesPerDay \times SizePerWrite \times RetentionDays$
    • •Example: 10k users, 2GB videos = 20TB storage? -> S3 Cold Storage needed.
  3. •High-Level Design (The "Blob" Phase)
    • •Client -> API Gateway -> Service -> DB.
    • •Validate against constraints (Will single DB handle calc QPS? No -> Read Replica).
  4. •Detailed Design (The "Hard Parts")
    • •"How exactly do we handle the video processing failure?" (Dead Letter Queue + Retry)
    • •"How do we handle 1 million users?" (Sharding vs Partitioning)

⚖️ Decision Matrix (Trade-off Guide)

StyleGood ForBad ForComplexities
MonolithSpeed, Simplicity, small/med teamsIndependent scaling, Large teamsTight coupling
MicroservicesIndependent scaling, Polyglot, 100+ devsComplexity, Latency, Data consistencyDistrib. Tracing, Eventual Consistency
ServerlessSpiky traffic, Low ops, Event-drivenLong-running tasks, Cold startsVendor lock-in, Debugging

📊 Capacity Planning Cheatsheet

  • •QPS to Servers: $Servers = TargetQPS / (SingleCoreQPS \times Cores \times UtilizationFactor)$
  • •Bandwidth: $Mbps = TotalBytesPerSec * 8 / 1,000,000$
  • •Database: Read-heavy? Cache/Replica. Write-heavy? Sharding/Queue-buffering.

Validation Checklist

Before finalizing architecture:

  • • Requirements clearly understood
  • • Constraints identified
  • • Each decision has trade-off analysis
  • • Simpler alternatives considered
  • • ADRs written for significant decisions
  • • Team expertise matches chosen patterns