convez.ai
SkillProdutividadeIntermediário
Royalti-io

Groundwork

Royalti-io/groundwork

Plano vivo e estruturado para projetos com re-execução segura

Ver no GitHub
BomPluginScore · de 0 a 100

Skill para Claude Code que scaffolda e mantém uma pasta de plano vivo. Cria documentação estruturada (pesquisa, design, orquestração) com blocos protegidos que nunca são sobrescritos em re-execuções. Oferece um board HTML interativo com visualizações Kanban e grafo de dependências.

Instalação

$npx convez-ai Royalti-io/groundwork

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
2
Forks
0
Acompanhando
0
Saúde
100%
Último commit
há 1 dia
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Planejar features de software com pesquisa e design options
  • Scaffoldar campanhas e mudanças organizacionais
  • Rastrear work packages com dependências e status
  • Documentar decisões e riscos de projeto com re-sincronização automática

Veredito

Skill maduro que resolve um problema real: manter planos vivos sem perder edições humanas. A abordagem com checksums para decidir re-escritas é sólida. Suporta múltiplos perfis (software, general, content, design-system), tornando-o versátil para diferentes contextos.

README

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

groundwork

A Claude Code skill that scaffolds and maintains a living plan folder. Re-runs augment your work — they never overwrite it.

npx skills add royalti-io/groundwork

Works with Claude Code, Codex, Gemini, Cursor, and 70+ other agents.


Quickstart demo

▶ Open interactive quickstart demo (init → research → review → orchestrate — 4 steps, keyboard-navigable)


What it does

You give groundwork a goal. It scaffolds a folder of numbered, living documents — a plan, two research files, a discussion log, a tracking file, and a standalone HTML board you can open in any browser. Then it gives you a small set of actions you can run at any time: research, design, review, orchestrate, refresh-board.

The part that matters: every block the skill generates is wrapped in a fenced region. Everything outside a fence is yours and is never touched. When you re-run an action, a checksum decides what changes on disk — not the model. A vibe you can't trust; a sha256 you can.


Plan board

The board reads your tracking file and renders three views: mission control, Kanban, and a dependency-graph of your wave plan.

groundwork board — Kanban view with status columns and a work-package brief panel

groundwork board — DAG view showing wave order and gate dependencies

It is a self-contained HTML artifact: open it in any browser, no server needed. Inside the Ikenga workspace it renders live next to your running sessions.


The folder shape

plans/your-feature/
├── .groundwork.json          ← identity + state anchor
├── 00-README.md              ← north star + links
├── 01-plan.md                ← goal, phases, architecture, risks
├── 02-research-external.md   ← prior art, competitors, libraries
├── 03-research-internal.md   ← codebase, schema, constraints
├── 04-discussion.md          ← review rounds, newest-first
├── 05-tracking.md            ← WPs, deps, DoDs, status
├── 09-orchestration.md       ← wave plan + per-WP kickoff briefs
└── artifact/
    └── board.html            ← standalone plan board

Folder treemap — the spine visualised as a proportional area map


Profiles

A profile swaps vocabulary and optional blocks — not the spine. The safe-regeneration machinery is identical across all four.

ProfileForWork unit
softwareFeatures, code workwork package / PR
generalCampaigns, org changes, non-codeworkstream / deliverable
contentEditorial work, content seriespiece / asset
design-systemComponent/token systemspart

The design-system profile adds a parts gallery, token pipeline, and a per-part quality gate.

design-system profile — living component gallery with foundations and parts tracked against tokens


Action set

ActionWhat it does
initInterview + scaffold the folder skeleton
researchFill 02/03 research files (external + internal)
designProduce ≥2 comparable design options, lock one
subplanScaffold a focused sub-plan (diff-plan / decision-doc / bug-doc)
reviewGap analysis → new Round in 04 → re-sync tracking
clarifyReadiness gate before orchestrate
orchestrateEmit 09-orchestration.md with wave plan + WP briefs
refresh-boardRegenerate artifact/board.html from current docs
statusRead-only freshness + ID + coverage report

Add --emit-workflow to orchestrate for a runnable Claude Code Workflow that fans waves out in parallel and turns freeze gates into sign-off barriers.


Install

npx skills (recommended)

# Global — available across all projects
npx skills add royalti-io/groundwork -g

# Project — committed with your repo, shared with the team
npx skills add royalti-io/groundwork

Git clone

git clone https://github.com/royalti-io/groundwork.git
cp -r groundwork/skills/groundwork ~/.claude/skills/

Curl one-liner

curl -sSL https://raw.githubusercontent.com/royalti-io/groundwork/main/install.sh | bash

The installer drops the skill into ~/.claude/skills/groundwork/ via symlink against a cached clone in ~/.cache/ikenga-skills/, so git pull is the update path.


Usage

After install, in any Claude Code session:

/groundwork init plans/<your-plan>/ --profile software --goal "…"

Then run actions as the work progresses:

/groundwork research plans/<your-plan>/
/groundwork review plans/<your-plan>/
/groundwork orchestrate plans/<your-plan>/

See skills/groundwork/SKILL.md for the full agent-facing spec.


Further reading


License

Apache-2.0. Copyright © 2026 Royalti.io.

Ver o original no GitHub

Bom para

Relacionados