Waza is a skill system that packages engineering habits into executable routines Claude AI agents can run.
The tool addresses the challenge of making AI agents more effective at complex engineering tasks by codifying proven practices into discrete, reusable skills. Rather than relying on general-purpose prompting, Waza bundles domain-specific workflows—such as systematic debugging, design pressure-testing, and code review—into named skills that agents invoke through slash commands or direct invocation. Each skill encodes a complete playbook: the `/think` skill challenges problems and produces implementation-ready plans; `/hunt` performs systematic debugging with root-cause confirmation; `/check` reviews diffs and handles release workflows; `/ui` drives frontend design through screenshot iteration; `/write` adapts prose for natural expression in Chinese and English; `/learn` structures research into a six-phase workflow; `/read` handles URL and PDF ingestion with platform-specific routing; and `/health` audits agent configuration and maintainability. The skills are installed as a single package and documented with reference materials, helper scripts, and failure patterns from real use.
Waza suits teams using Claude Code or Codex who want to standardize how AI agents approach recurring engineering tasks. It works best for projects where consistency in code review, debugging methodology, and design iteration matters—teams that have developed strong engineering practices and want those practices to guide AI agent behavior. The tool is particularly valuable when you need agents to handle domain-specific constraints and follow project-specific conventions rather than applying generic defaults.
The maintainers respond to issues and pull requests within a day. Development work concentrates on bug fixes and documentation improvements.
