---

name: reviewing-readability description: > Code readability review based on "The Art of Readable Code" and Miller's Law (7±2). Triggers: 可読性, 理解しやすい, わかりやすい, 明確, 命名, 変数名, 関数名, ネスト, 深いネスト, 関数設計, コメント, 複雑, 難しい, 難読, Miller's Law, ミラーの法則, 認知負荷, AI-generated, 過剰設計. allowed-tools:

• Read
• Grep
• Glob
• Task agent: readability-reviewer user-invocable: false

---

Readability Review - Code Clarity & Cognitive Load

Target: New team member understands code in < 1 minute.

Miller's Law Limits (7±2)

Context	Ideal	Maximum
Function arguments	3	5
Class methods	5	7
Conditional branches	3	5
Function length	5-10 lines	15 lines
Nesting depth	2	3

Section-Based Loading

Section	File	Focus	Triggers
Control Flow	references/control-flow.md	Nesting, guard clauses	ネスト, Miller's Law
Comments	references/comments-clarity.md	Why not What, intent	コメント, 意図
AI Antipatterns	references/ai-antipatterns.md	Over-engineering detection	AI-generated, 過剰設計

Note: Naming conventions → @../../patterns/code/naming.md

Quick Checklist

Naming

• Concrete over abstract (validateUserEmail not processData)
• Searchable, pronounceable names
• Intent is obvious from name

Control Flow

• Nesting depth ≤ 3
• Guard clauses for early returns
• Complex conditions extracted to functions

Comments

• Explain "why", not "what"
• Update or delete outdated comments
• Code is self-documenting first

AI Code Smells

• No premature abstractions (interface for single impl)
• No unnecessary classes for simple tasks
• No "future-proof" flexibility nobody asked for

Key Principles

Principle	Application
Clarity > Cleverness	Simple code wins
Respect 7±2 limit	Break into chunks
Tell, Don't Ask	Direct method calls

References

• [@../../skills/applying-code-principles/SKILL.md] - Cognitive limits science
• [@../../rules/development/READABLE_CODE.md] - Detailed guidelines
• @./references/ai-antipatterns.md - AI over-engineering patterns