メニュー
Produces a system design doc for a given requirement. Use when the user asks to "design a system", "architect a service", "design X at scale", or invokes /sdd:design-system. Grounded in DDIA and Fundamentals of Software Architecture.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | design-system |
| description | Produces a system design doc for a given requirement. Use when the user asks to "design a system", "architect a service", "design X at scale", or invokes /sdd:design-system. Grounded in DDIA and Fundamentals of Software Architecture. |
| argument-hint | <requirement, e.g. 'URL shortener' or 'multi-tenant chat'> |
| allowed-tools | Read, Write, Grep, Glob |
Produce a design doc a staff engineer would respect — opinionated, grounded, quantitative.
ASSUMPTION:. Do not pause to ask — iterate.references/arch-styles.md for topologies.latency-numbers.md, capacity-estimation.md.design-database skill.| Topic | File |
|---|---|
| Architecture style | arch-styles.md |
| Capacity / latency | capacity-estimation.md, latency-numbers.md |
| Caching | caching.md |
| Load balancing / rate limiting | load-balancing.md, rate-limiting.md |
| Queues / streams | messaging.md |
| APIs | api-design.md |
| Resilience | resilience.md |
| Observability | observability.md |
| Security | security.md |
| Deployment | deployment.md |
| Microservices patterns | microservices-patterns.md |
| Real-time / WebSocket / SSE | real-time-systems.md |
| ETL / CDC / lakehouse | data-engineering.md |
| ML / RAG / vector DB | ml-ai-serving.md |
Load at most 2–3 per question. Do not pre-load.
# System Design: <name>
## 1. Requirements (functional / non-functional / assumptions)
## 2. Capacity estimates (numbers, not prose)
## 3. High-level architecture (component list + Mermaid)
## 4. Data model (store, sharding, replication, cache)
## 5. API (endpoints or RPCs)
## 6. Async & real-time
## 7. Resilience & failure modes
## 8. Observability
## 9. Security
## 10. Deployment
## 11. Trade-offs
Offer to save to ./designs/<kebab-name>.md. Offer to generate diagrams via diagram.
Generates a complete, production-ready UI from a natural-language prompt — React + Tailwind, grounded in all 8 UIX reference books + WCAG 2.2, ROIN baseline always on. Asks 4 targeted questions then writes a self-contained .tsx file and compiles it to a browser-openable preview.html. Auto-activates on "design a [X] from scratch", "build me a UI for [X]", "create a full [X] page", "generate a UI for [X]", or /uix:generate-design.
This skill should be used when the user asks to "analyze this file", "check this code", "scan for violations", "find clean code issues", "review code quality", "check file length", "audit this project", "what's wrong with this code", or when a file looks messy or overly long. Scans files for clean code violations and outputs a grouped report of critical issues, warnings, and style suggestions.
This skill should be used when the user asks to "refactor this file", "clean up this code", "extract this function", "break this function apart", "pull this out into a method", "simplify this switch", "turn this into a table", "inline this", "replace magic number", or mentions a specific refactoring by name. Applies one named, surgical refactoring from the textbook catalog (Extract Method, Table-Driven, Replace Conditional with Polymorphism, etc.) and shows exactly what changed.
This skill should be used when the user asks to "clean my code", "make it human readable", "make this readable", "rewrite this cleanly", "refactor this", "clean up this function", "simplify this code", "make this code better", "fix code quality", "clean this file", "this code is messy", "improve this code", or when code has visible violations like long functions, deep nesting, poor names, or missing interfaces. Produces a clean version preserving all behavior, with reasoning for each change.
This skill should be used when the user asks to "check error handling", "find silent failures", "fix silent errors", "add input checks", "fail fast", "make this safer", "check for hidden errors", "defensive programming", or mentions empty catch blocks, swallowed exceptions, or missing validation. Detects and optionally fixes Rule 12 violations — silent error swallowing and missing guard clauses.
This skill should be used when the user asks to "suggest a better structure", "apply a design pattern", "this switch is too big", "simplify this class", "apply strategy pattern", "apply command pattern", "apply factory pattern", "fix this structure", "replace switch with polymorphism", or mentions restructuring a class or giant switch statement. Detects and optionally applies well-known design patterns (Strategy, Command, Factory, State) from OOP vs. Functional Programming.