tw93/Waza

What it solves

Waza provides a set of structured engineering habits—turned into "skills"—that prevent AI agents from producing generic or imprecise work. It transforms raw AI output into professional engineering workflows by enforcing specific outcomes, constraints, and verification steps for common tasks like planning, debugging, and reviewing.

How it works

It functions as a collection of installable skills for AI coding agents (such as Claude Code, Codex, Cursor, and others). Each skill is a specialized playbook that the agent follows when triggered by a slash command or name. These skills can be carried out in sequence (chaining) to complete complex workflows, such as using /think to plan a feature and /check to verify it before merging.

Who it’s for

Software engineers who use AI agents to write code and want to move beyond basic prompting to implement professional, systematic engineering habits (e.g., root-cause debugging and pressure-testing designs).

Highlights

• Eight Core Skills: Includes specialized workflows for thinking/planning (/think), UI design (/ui), verification (/check), root-cause debugging (/hunt), prose polishing (/write), research (/learn), reading documents (/read), and agent health auditing (/health).
• Cross-Agent Compatibility: Works across multiple platforms including Claude Code, Codex, Cursor, Kimi Code CLI, and Claude Desktop.
• Manual Chaining: Allows users to manually trigger transitions between skills to maintain control over the AI's process.
• Project-Awareness: The /check skill dynamically reads public repository context (like READMEs and manifests) to ensure reviews are project-specific.
• Optional Guardrails: Provides optional rules for English coaching, anti-pattern prevention, and routing hints.