convez.ai
SkillProdutividadeIntermediário
darkroomengineering

CC Settings — Configuração Claude Code da Darkroom

darkroomengineering/cc-settings

One-command setup de agentes, skills e padrões para Claude Code

Ver no GitHub
BomPluginScore · de 0 a 100

Skill que instala agentes, skills, hooks e padrões de código no Claude Code com um comando. Oferece 9 subagentes especializados, 34 skills auto-invocáveis e composição modular de settings sem sobrescrever configurações locais. Inclui detecção de adulteração e suporte a múltiplos perfis de stack.

Instalação

$npx convez-ai darkroomengineering/cc-settings

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
54
Forks
6
Acompanhando
0
Saúde
62%
Último commit
há 1 dia
Versão
v16.4.0
Issues abertas
0
Contribuidores
5

Casos de uso

  • Sincronizar configuração de Claude Code em equipes
  • Instalar subagentes e skills especializados para workflows
  • Gerenciar hooks, MCP servers e permissões de forma modular
  • Migrar entre diferentes stacks (Next.js, React Native, Tauri, etc)

Veredito

Ferramenta bem arquitetada para equipes que usam Claude Code, com suporte a rollback e instalação não-destrutiva. Reduz atraso na configuração compartilhada, mas requer Bun e restringe-se ao ecossistema Claude Code.

README

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

cc-settings

Claude Code configuration for the Darkroom team — installs agents, skills, hooks, and coding standards into ~/.claude/.


Features

  • One-command install — drops shared agents, skills, hooks, and standards into ~/.claude
  • Subagents & skills — nine specialized subagents and a curated, auto-invocable skill library
  • Composed settings — settings.json assembled from modular config: permissions, hooks, and MCP servers
  • Non-destructive — existing permissions, custom hooks, and local overrides survive re-installs
  • One-command rollback — restore the previous backup if anything looks off
  • Tamper detection — a fingerprint plus audit guard the hooks against supply-chain attacks
  • Open standard — AGENTS.md is read by Codex, Cursor, Copilot, and Windsurf too

Install

macOS / Linux:

bash <(curl -fsSL https://raw.githubusercontent.com/darkroomengineering/cc-settings/main/setup.sh)

Windows (PowerShell):

powershell -ExecutionPolicy Bypass -c "irm https://raw.githubusercontent.com/darkroomengineering/cc-settings/main/setup.ps1 | iex"

Both one-liners clone the repo and run the installer. To pass flags (--light, --dry-run, …), clone first and run the bootstrap from the checkout: bash setup.sh --light or .\setup.ps1 --light.

Requires Bun ≥ 1.1.30 and git — the bootstrap installs Bun automatically if missing. Re-installs are non-destructive: existing permissions, custom hooks, and local overrides survive.

Restart Claude Code after install.

Light profile (for newcomers)

New to Claude Code and don't want the full surface? Install the light profile — raw Claude Code with only two additions: the statusline and the share-learning skill.

bash setup.sh --light    # macOS / Linux
.\setup.ps1 --light      # Windows (PowerShell)

No custom CLAUDE.md, agents, rules, profiles, MCP servers, hooks (beyond the statusline), or effort overrides — just vanilla Claude Code so you're not overwhelmed. Re-run bash setup.sh without --light any time to upgrade to the full config; both tiers are permanently supported. See MANUAL.md for the full comparison.


What gets installed

~/.claude/
├── AGENTS.md           # Portable coding standards (read by all AI tools)
├── CLAUDE.md           # Claude-Code-specific config
├── settings.json       # Composed from config/*.json (permissions, hooks, MCP)
├── agents/             # 9 specialized subagents
├── skills/             # 34 auto-invocable skills
├── profiles/           # Stack contexts: nextjs, react-native, tauri, webgl, maestro, react-router
├── rules/              # Path-conditioned rules (load on-demand by file type)
└── src/                # Hook + script implementations (TypeScript)

Repo → install mapping:

Repo dirInstalls to
agents/~/.claude/agents/
skills/~/.claude/skills/
rules/~/.claude/rules/
profiles/~/.claude/profiles/
config/*.json~/.claude/settings.json (composed)

Common commands

bash setup.sh                  # Install / update (full profile)
bash setup.sh --light          # Light profile: raw CC + statusline + share-learning only
bun src/setup.ts --rollback    # Restore the previous backup
bun src/setup.ts --dry-run     # Preview what would change
bun run compose                # Preview the composed settings.json
bun run lint:skills            # Validate skill frontmatter + cap (≤40)
bun run audit:hooks            # Classify hooks as trusted/unknown/suspicious
bun run typecheck              # TypeScript check
bun test                       # Run all tests

Docs

DocWhat's in it
MANUAL.mdEvery skill — how to invoke it, what it does
AGENTS.mdCoding standards and guardrails (source of truth)
CLAUDE.mdClaude-Code config, delegation rules, effort levels
docs/profiles.mdStack-specific profiles (Next.js, RN, Tauri, WebGL, React Router)
docs/skill-authoring.mdWriting new skills
CHANGELOG.mdRelease history

Why

Shared config means every engineer on the team runs the same agents, the same guardrails, and the same coding standards — no per-machine drift. AGENTS.md is the open standard for AI coding instructions and is also read by Codex, Cursor, Copilot, and Windsurf, so the rules follow you across tools.


darkroom.engineering · MIT

Ver o original no GitHub

Bom para

Relacionados