Waza

Waza is a collection of eight skills for Claude Code (Anthropic's AI coding assistant) that encode common software engineering habits as slash commands you can invoke during your work. The name comes from a Japanese martial arts term for a practiced technique: something drilled until it becomes instinct. The idea is that good engineering involves more than writing code, and that an AI assistant can handle each part of the workflow if given the right structure.

The eight skills each target a different moment in the development process. Before building something new, /think challenges the requirements and produces a decision-ready plan. During frontend work, /design generates UI with a committed visual direction. After finishing a task, /check reviews the diff and verifies the work before merging or releasing. When debugging, /hunt walks through the problem systematically and confirms the root cause before touching anything. Other skills cover writing clear prose (/write), researching an unfamiliar topic (/learn), fetching and cleaning web content or PDFs (/read), and auditing your Claude Code configuration (/health).

Skills can be chained together manually. A typical flow might run /think to plan, then implement the feature, then /check before merging. The README provides several common workflow patterns for design, debugging, research, and shipping fixes.

Installation is one command via npx for Claude Code or Codex. A Claude Desktop ZIP package is also available. The skills install globally so they are available in any project.

Waza is part of a three-project set from the same author: Kaku for code generation and Kami for document output. Two optional extras are also included: a statusline showing context window and quota usage, and an English coaching rule that quietly corrects grammar mistakes in your prompts. The project is MIT licensed.