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, committing at each step, so improvements are safe, incremental, and reversible.

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