/ Directory / Playground / Waza (技)
● Community tw93 ⚡ Instant

Waza (技)

by tw93 · tw93/Waza

Senior-engineer habits packaged as Claude Code skills — small commits, tight refactors, focused debugging, deliberate reviews. Stuff you know, made repeatable.

Waza (Japanese 技 — 'technique') is a curated set of engineering-discipline skills by tw93, a maintainer at the eSheep / Pake / Waza ecosystem. Each skill encodes a specific habit: 'small atomic commit', 'red-green-refactor', 'systematic bisect', 'code review with checklist'. They're not flashy features — they're the small disciplines that separate a senior dev's output from a junior's, applied automatically.

Why use it

Key features

Live Demo

What it looks like in practice

ready

Install

Author repository · Claude Skills documentation

Use Cases

Real-world ways to use Waza (技)

Make Claude write small, atomic commits instead of one giant blob

👤 Anyone whose Claude PRs end up as 1 commit with 600 changed lines ⏱ ~30 min intermediate

When to use: PR review is hell because the commit is unreviewable.

Flow
  1. Plan commits
    Use waza commit skill. Plan the work as a sequence of atomic commits before writing any code.✓ Copied
    → Numbered commit plan; each one is a one-line summary
  2. Execute one at a time
    Make commit 1. Stop. Show me the diff and message.✓ Copied
    → Single small commit; message follows convention
  3. Continue or revise
    Continue with the next commit, or revise the plan if step 1 changed your understanding.✓ Copied
    → Either next commit or revised plan

Outcome: PRs you can actually review.

Pitfalls
  • Plan is too granular (10 commits for trivial change) — Skill default targets 3–7 commits per PR; pass --target N to override
Combine with: github

Apply red-green-refactor on a refactor that's worth doing

👤 Devs refactoring legacy code ⏱ ~90 min intermediate

When to use: You want to refactor; you also don't want to break things.

Flow
  1. Cover with tests
    Use waza refactor skill on /src/<module>.ts. First step: characterize current behavior with tests.✓ Copied
    → Tests for current behavior; passing
  2. Refactor
    Now refactor incrementally. Tests must stay green at every step.✓ Copied
    → Series of refactor commits, all green

Outcome: Cleaner code with proof it didn't break.

Pitfalls
  • 'Tests for current behavior' includes existing bugs — Skill flags suspect-looking tests; mark and refactor those separately

Run a structured code review on a PR

👤 Reviewers who want consistency ⏱ ~30 min beginner

When to use: Reviewing a PR you didn't write — easy to miss things.

Flow
  1. Walk the diff
    Use waza review skill on PR #123. Walk the diff hunk by hunk; check: correctness, naming, error handling, tests, security.✓ Copied
    → Per-hunk notes; any issue surfaced
  2. Summary
    Write the review summary — 3 strengths, 3 concerns, 1 question.✓ Copied
    → Posted-style summary

Outcome: Reviews that find things, not 'lgtm' theater.

Pitfalls
  • Skill nitpicks style when bigger issues exist — Reorders by severity; trust the order
Combine with: github

Combinations

Pair with other MCPs for X10 leverage

waza-skill + github

Pair with GitHub MCP for end-to-end PR lifecycle

Plan, commit, push, open PR, request review — chain waza commit + github.✓ Copied
waza-skill + ccpm-skill

CCPM owns the issue lifecycle, Waza owns the engineering discipline within

/start issue 42, then run waza commit.plan inside the worktree.✓ Copied

Tools

What this MCP exposes

ToolInputsWhen to callCost
commit.plan task_description Start of any non-trivial work 0
refactor.tdd path, target Refactor with safety net 0
debug.systematic symptom, context Bug that resists guessing 0
review.checklist pr_url Reviewing a PR 0

Cost & Limits

What this costs to run

API quota
N/A
Tokens per call
Light — skills are short markdown
Monetary
Free
Tip
Habits compound — apply consistently rather than only when stuck

Security

Permissions, secrets, blast radius

Credential storage: None
Data egress: Local

Troubleshooting

Common errors and fixes

Skill keeps splitting commits when single is fine

Smallest unit-of-change varies; pass an example PR for calibration

Refactor tests cover too little

Skill warns when coverage <30% on the touched module — add manual seeds

Alternatives

Waza (技) vs others

AlternativeWhen to use it insteadTradeoff
wshobson/agentsYou want role-based agents (frontend, backend) rather than habit skillsDifferent shape; both are fine — different abstraction levels
addyosmani/agent-skillsYou want web-perf/a11y depth specificallyAddy's set is web-frontend; Waza is general engineering

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills

Continue exploring

Related tools and workflows

Choose another guide for your workflow, or explore a collection.

Browse all topic collections → · See 29 new official additions →