convez.ai
MCP ServerAutomaçãoAvançado
josh-c2cc

Claude OS Starter

josh-c2cc/claude-os-starter

Sistema operacional completo para Claude Code com workflows, plugins e memória compartilhada

Ver no GitHub
InicialPluginScore · de 0 a 100

Kit de inicialização para transformar Claude Code em um sistema operacional produtivo com workflows estruturados, revisão cruzada entre modelos (Codex + Gemini), plugins, conectores MCP e memória persistente entre sessões. Oferece um modelo mental em 6 camadas com configuração pronta para bootstrap.

Instalação

$npx convez-ai josh-c2cc/claude-os-starter

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
1
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Estruturar workflows profissionais no Claude Code com skills reutilizáveis
  • Implementar revisão cruzada de código entre múltiplos modelos de IA
  • Integrar MCP connectors para automatizar tarefas externas (deploy, geração de imagens, calendários)
  • Construir segunda cérebro persistente que mantém contexto entre sessões

Veredito

Um kit arquitetural sólido para quem quer elevar Claude Code além do básico, com scaffolding bem documentado e modelo mental claro em 6 camadas. Requer disposição para copiar, customizar e manter a configuração, mas oferece estrutura séria para workflows profissionais em larga escala.

README

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

claude-os-starter

A starter kit for turning Claude Code into a real operating system for your work: a disciplined workflow (skills), cross-model review (Codex + Gemini), a plugin loadout, opt-in MCP connectors, and a persistent second-brain memory that survives across sessions and machines.

It's a reference setup you bootstrap from, not a framework you install. You copy the pieces you want into your own ~/.claude, follow a guided checklist, and end with a power-user Claude Code that reviews its own work and remembers what matters.

New here? Go straight to SETUP.md — a copy-paste-exact, Windows-first walkthrough (with macOS/Linux notes) that takes you from zero to a verified setup.

What this is / isn't

  • Is: generic, reusable scaffolding — behavioral rules, ~12 workflow skills, peer-agent review, a plugin/MCP/memory setup, and a guided install.
  • Isn't: anyone's private config. No credentials, no client data. Sections marked EXAMPLE (a stack, region defaults, a skip-list) are starting points to replace with your own.

The 6-layer mental model

Your "operating system" is really six layers that live in different places and install differently. SETUP.md walks them in practical order.

#LayerWhat it isWhere it lives
0Claude CodeThe CLI itself (prerequisite)global npm install
1Core configCLAUDE.md (behavioral rules), settings, skills, commands, hooks, an agent~/.claude/
2PluginsBundled skills/agents from marketplaces (superpowers, code-review, …)installed via /plugin
3MCP connectorsTools that act on external services (image gen, deploy, calendar…)account authorizations
4Peer agentsCodex + Gemini CLIs for cross-model review & multimodal/large-contexton your PATH
5Second brainFile-based memory + handoffs that persist across sessions/machinesyour vault ($CLAUDE_VAULT)

(Your project repos are where the work happens — out of scope for the kit, but layers 1–5 serve them.)

What's in the box

claude/                  # maps into ~/.claude/
  CLAUDE.md.template        # 18 behavioral rules (verbatim) + stubbed profile/stack
  settings.json.example     # allowlisted settings + a .md explaining each key
  statusline.ps1            # model · context% · git branch · dir
  hooks/                    # biome auto-format + cross-device task mirror
  commands/                 # /codex-review, /gemini-analyze, /gemini-codebase-qa
  agents/                   # web-performance-auditor
  skills/                   # 12 workflow skills (begin, end, handoff, resume,
                            #   peer-agent-routing, context-budget, doubt-driven-development,
                            #   ai-regression-testing, observability-and-instrumentation,
                            #   security-and-hardening, source-driven-development, notebooklm)
vault/                   # second-brain scaffold (MEMORY.md template + README)
docs/                    # plugins.md · mcp-connectors.md · peer-agents.md
scripts/                 # verify-setup.ps1 / .sh
SETUP.md                 # the guided install

Quickstart (5 minutes)

  1. Install Claude Code (layer 0).
  2. Add the marketplaces and install the core plugins (layer 2).
  3. Copy claude/ into ~/.claude/ (back up any existing config first), rename CLAUDE.md.template → CLAUDE.md, fill the placeholders (layer 1).
  4. Install Codex + Gemini CLIs (layer 4).
  5. Point $CLAUDE_VAULT at a vault folder and create MEMORY.md (layer 5).
  6. Run scripts/verify-setup.ps1 (or .sh) — aim for all PASS.

Each step, with exact commands and "what you should see," is in SETUP.md. MCP connectors (layer 3) are entirely optional — add them later from docs/mcp-connectors.md.

License

MIT — see LICENSE.

Ver o original no GitHub

Bom para

Relacionados