convez.ai
SkillProdutividadeAvançado
Nikitzu

mzyx-skills

Nikitzu/mzyx-skills

Pipeline estruturado de engenharia: brainstorm → spec → plan → build → ship

Ver no GitHub
InicialPluginScore · de 0 a 100

Conjunto curado de skills para Claude Code que integra três bibliotecas upstream (superpowers, agent-skills, waza) em um pipeline de engenharia diária. Cobre camadas de processo (brainstorm → plan → execute → verify → review), domínio (TDD, debugging, performance) e disciplina (think, design, check, hunt, research).

Instalação

$npx convez-ai Nikitzu/mzyx-skills

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 1 dia
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Estruturar fluxo de trabalho de engenharia com comandos slash (brainstorm, spec, plan, build, ship)
  • Implementar TDD e verificação automática antes de completar tarefas
  • Conduzir revisões de código, auditorias de segurança e testes em paralelo durante deploy
  • Root-cause debugging hipótese-first com skill 'hunt'
  • Pesquisa estruturada em seis fases e auditoria de configuração do Claude Code

Veredito

Excelente para times de engenharia que buscam automatizar pipeline estruturado. Sem duplicação desnecessária das upstream, mas requer familiaridade com Claude Code para aproveitar toda potência. Recomendado para fluxos rigorosos (TDD, review-gate, ship-pipeline).

README

A tradução para o português deste README ainda está na fila.

mzyx-skills

A curated, trimmed mix of three upstream Claude Code skill libraries — superpowers, agent-skills, and waza — assembled into one plugin tuned for daily engineering use.

Why

Both upstreams are excellent, but installing both side by side produces heavy duplication (review/plan/test all appear in both, plus alias commands and deprecated shims). This repo cherry-picks one canonical skill per concern and drops everything covered by personal rules files.

Three-layer model:

  • Process layer (superpowers). How work flows. Brainstorm → plan → execute → verify → review → finish. Lifecycle scaffolding.
  • Domain layer (agent-skills). How to do the work in each phase. TDD, debugging, simplification, performance, shipping. Content checklists.
  • Discipline layer (waza). Sharper opinions on think/design/check/hunt plus standalone utilities for research (/learn), URL fetch (/read), prose polish (/write), and config audit (/health). Eight skills, each a single habit with a clear trigger.

Slash command flow

The pipeline. Each command invokes one or more skills. Commands ship in .claude/commands/; specialist agents in agents/.

/brainstorm  →  /spec  →  /plan  →  /build  →  /ship
   (sp)         (as+sp)   (sp)     (asks!)    (3 agents+sp)
CommandSkills / agents invokedSource
/brainstormmzyx-skills:brainstormingsp
/specmzyx-skills:brainstorming (if fuzzy) → mzyx-skills:spec-driven-developmentsp + as
/planmzyx-skills:writing-planssp
/buildAsks the user which mode first: parallel teammates / separate session / inline solo. Each mode pairs mzyx-skills:test-driven-development + mzyx-skills:verification-before-completion per task; the parallel mode layers mzyx-skills:using-teams on top.sp + as
/shipParallel fan-out: code-reviewer + security-auditor + test-engineer agents → mzyx-skills:verification-before-completion → GO / NO-GO + rollback3 agents + sp

Review, TDD, and simplification flow through skill auto-trigger on natural language ("review my changes", "write a failing test", "simplify this") — no wrapper commands needed.

(sp = superpowers, as = agent-skills.)

Discipline-layer skills (waza) ship without command stubs — invoke directly via the namespaced skill name or rely on auto-trigger:

SkillWhen
mzyx-skills:thinkrough idea → decision-complete plan before code
mzyx-skills:designUI direction-lock + screenshot iteration
mzyx-skills:checksolo diff review + verification (lighter than /ship)
mzyx-skills:hunthypothesis-first root-cause debugging
mzyx-skills:learnsix-phase research → publish workflow
mzyx-skills:readfetch URL/PDF as clean Markdown
mzyx-skills:writestrip AI-pattern prose
mzyx-skills:healthClaude Code config-stack audit

Skills also auto-trigger on natural language as in the original repos — e.g. saying "fix this bug" still fires hunt. Commands are explicit verbs for the structured pipeline, not the only entry point.

Lifecycle map

StageSkillSource
Session startusing-skillssuperpowers
Explore intentbrainstormingsuperpowers
Spec (multi-file, ambiguous)spec-driven-developmentagent-skills
Plan (clear-enough)writing-planssuperpowers
Implement (this session)subagent-driven-developmentsuperpowers
Implement (separate session)executing-planssuperpowers
Write code (logic / bug / change)test-driven-developmentagent-skills
Hit a bug / unexpected errorhuntwaza
UI / frontendfrontend-ui-engineeringagent-skills
Refactor / clean upcode-simplificationagent-skills
Pre-claim doneverification-before-completionsuperpowers
Self-reviewcode-review-and-qualityagent-skills
Perf concernperformance-optimizationagent-skills
Record decisiondocumentation-and-adrsagent-skills
Write / edit a skillwriting-skillssuperpowers
Plan / pressure-test designthinkwaza
UI direction lockdesignwaza
Solo review + verifycheckwaza
Hypothesis-first debughuntwaza
Polish prosewritewaza
Multi-source researchlearnwaza
Fetch URL / PDFreadwaza
Audit Claude Code confighealthwaza

Install

Recommended — install as a plugin

This is the right install path. All skill cross-references in this repo use the mzyx-skills: prefix, which only resolves correctly when the plugin is loaded by name.

/plugin marketplace add github.com/Nikitzu/mzyx-skills
/plugin install mzyx-skills@mzyx-skills

After install, skills appear as mzyx-skills:brainstorming, mzyx-skills:test-driven-development, etc. in the listing, and these slash commands become available:

  • Pipeline commands (sp/as): /brainstorm /spec /plan /build /ship
  • Discipline skills (wz): mzyx-skills:think · design · check · hunt · write · learn · read · health (no command stubs — invoke namespaced or rely on auto-trigger)

Alternative — symlink into ~/.claude/skills/

Skills only, no commands. Bare names (no mzyx-skills: prefix) — cross-references inside skills will fail to resolve unless the prefix matches your install path, so prefer the plugin install above.

git clone https://github.com/Nikitzu/mzyx-skills.git ~/Code/mzyx-skills
ln -s ~/Code/mzyx-skills/skills/* ~/.claude/skills/

Updating from upstream

This repo is a curated fork. To pull selective updates from upstream:

git remote add upstream-superpowers https://github.com/obra/superpowers.git
git remote add upstream-agent-skills https://github.com/addyosmani/agent-skills.git
git remote add upstream-waza https://github.com/tw93/waza.git

git fetch upstream-superpowers
git fetch upstream-agent-skills
git fetch upstream-waza

# See what changed in skills you actually use
git log --oneline HEAD..upstream-superpowers/main -- skills/
git log --oneline HEAD..upstream-agent-skills/main -- skills/
git log --oneline HEAD..upstream-waza/main -- skills/

# Cherry-pick selectively (one file at a time keeps drift minimal)
git checkout upstream-superpowers/main -- skills/brainstorming/SKILL.md
git checkout upstream-agent-skills/main -- skills/test-driven-development/SKILL.md

Cherry-pick over merge. A merge would drag in the skills and behaviours explicitly removed here.

Waza re-strip on update. After cherry-picking from upstream-waza, re-run the strip pass to remove the 🥷 emoji preamble and Chinese trigger phrases before committing — the original waza skills carry both, and they have no purpose in this fork.

Excluded — and why

See NOTICE for the full exclusion list.

License

MIT — see LICENSE. All upstream attributions preserved.

Ver o original no GitHub

Bom para

Relacionados