Evolutionary Naming
kawasima/evolutionary-naming
Refine nomes incrementalmente através de 7 passos organizados em 3 fases
Ver no GitHubSkill para Claude Code que melhora nomes de identificadores através de refinamento progressivo em 3 fases (auditoria e melhoria). Caminha código por 7 passos evolutivos, gerando propostas de renomeação com mensagens de commit, mantendo mudanças seguras e reversíveis.
Instalação
npx convez-ai kawasima/evolutionary-naming1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Auditoria de qualidade de nomenclatura em bases de código existentes
- Refatoração incremental de métodos e classes com nomes pouco descritivos
- Melhoria estrutural e de design através de renomeação progressiva
Veredito

Ferramenta sólida e bem documentada para quem leva nomenclatura a sério. O modelo de 7 passos em 3 fases é fundamentado em trabalho acadêmico (Arlo Belshee) e a abordagem advisory (sem editar automaticamente) oferece controle total ao usuário. Ideal para times que buscam melhorar legibilidade de código de forma estruturada.
README
A tradução para o português deste README ainda está na fila.
evolutionary-naming
A Claude Code skill for evolutionary naming — improving names step-by-step through progressive refinement instead of trying to find the perfect name in one shot.
Naming is a process, not a single step. This skill walks names through 7 progressive steps grouped into 3 phases, with one commit per step, so improvements are safe, incremental, and reversible.
The skill is advisory: it reads code (Read/Grep/Glob) and proposes each rename plus the commit message you'd use — it doesn't edit files or run git. You apply the changes.
What it does
| Phase | Steps | Nature |
|---|---|---|
| Phase 1: Insight to Name | Missing → Nonsense → Honest → Honest and Complete | Universal. Pure naming. Safe. |
| Phase 2: Name to Structure | Honest and Complete → Does the Right Thing | Structural refactoring. Asks permission. |
| Phase 3: Combine for Design | Does the Right Thing → Intent → Domain Abstraction | Domain-level. Asks permission. |
The skill operates in two modes:
- audit-mode — exhaustive scan of provided code, classifies every identifier by its current step, outputs a structured table grouped by phase. Does not rename.
- improve-mode — interactive single-target improvement. Walks Phase 1 continuously, pauses before Phase 2 and Phase 3.
Match the depth of work to the user's actual story. Most naming improvements should stop after Phase 1.
Installation
Via marketplace (recommended)
/plugin marketplace add kawasima/evolutionary-naming
/plugin install evolutionary-naming@kawasima-skills
Via skills.sh
npx skills add kawasima/evolutionary-naming
Manual install (single user)
git clone https://github.com/kawasima/evolutionary-naming.git
cp -r evolutionary-naming/skills/evolutionary-naming ~/.claude/skills/
Claude Code auto-discovers skills from ~/.claude/skills/. No configuration required.
Usage
The skill activates automatically when you ask Claude to improve naming, refactor a class/method, or review naming quality.
Audit mode triggers:
- "このコードの命名で改善余地ある?"
- "命名レビューして"
- "naming audit"
Improve mode triggers:
- "
dを改善して" - "DocumentManager をリファクタリング"
- "this method's name is unclear"
When the request is ambiguous (large code paste with no specific target), the skill asks which mode to use.
Files
skills/evolutionary-naming/
├── SKILL.md router and overview
├── reference.md shared 7-step diagnosis, transitions, red flags
├── audit-mode.md exhaustive scan workflow + table output
└── improve-mode.md interactive single-target + pause protocol
License
CC BY 4.0. See LICENSE.
Background and Acknowledgments
The 7-step / 3-phase model used in this skill is based on Arlo Belshee's "Naming as a Process" series (CC BY 3.0), one of the most thorough public treatments of evolutionary naming. The skill adapts that material into Claude Code's skill format with audit/improve modes and phase-boundary safety gates.
References:
- Original blog series: https://arlobelshee.com/tag/naming-is-a-process
- Updated edition: https://www.digdeeproots.com/articles/naming-process/
Other influences worth exploring:
- Refactoring catalogs (extract method, rename, introduce parameter object) — Martin Fowler
- Domain-Driven Design (Whole Value, Ubiquitous Language) — Eric Evans