ganbatte-os 0.2.1

package/integrations/claude/README.md

@@ -0,0 +1,35 @@
+# Claude Adapter (Thin Integration)
+
+Este adapter define como o Claude Code consome artefatos do core G-OS.
+
+## Regras
+
+- Não duplicar conteúdo de negócio em `.claude/` (prompts, rules, skills, playbooks).
+- Sempre apontar para artefatos da raiz.
+- Manter em `.claude/` apenas:
+  - configurações locais da IDE;
+  - hooks e utilitários estritamente específicos da IDE;
+  - aliases/comandos que encaminham para o core.
+
+## Fonte de mapeamento
+
+- `command-map.json` é o contrato de mapeamento Claude -> comandos canônicos G-OS.
+- `agent-map.json` mapeia agentes canônicos para wrappers de comando do Claude.
+- `integrations/registry.json` mantém o registro global multi-IDE.
+- `agents/profiles/` contém os perfis de agente canônicos.
+- `.claude/commands/gos/agents/` contém wrappers adapter para Claude.
+- `mcp-specifics.md` contém regras de MCP específicas do Claude em camada de integração.
+
+## Sync de adapters
+
+```bash
+npm run sync:ides
+npm run check:ides
+```
+
+## Caminhos preferenciais
+
+- Prompts: `prompts/*.md`
+- Regras: `rules/*.md`
+- Agentes: `agents/.base/system-prompts/*.md`
+- Runtime no projeto alvo: `./.G-OS/*`

package/integrations/claude/agent-map.json

@@ -0,0 +1,46 @@
+{
+  "ide": "claude",
+  "version": "1.0.0",
+  "agentAdapters": [
+    {
+      "agentId": "ux-design-expert",
+      "canonicalProfile": "agents/profiles/ux-design-expert.md",
+      "adapterPath": ".claude/commands/gos/agents/ux-design-expert.md"
+    },
+    {
+      "agentId": "architect",
+      "canonicalProfile": "agents/profiles/architect.md",
+      "adapterPath": ".claude/commands/gos/agents/architect.md"
+    },
+    {
+      "agentId": "dev",
+      "canonicalProfile": "agents/profiles/dev.md",
+      "adapterPath": ".claude/commands/gos/agents/dev.md"
+    },
+    {
+      "agentId": "sm",
+      "canonicalProfile": "agents/profiles/sm.md",
+      "adapterPath": ".claude/commands/gos/agents/sm.md"
+    },
+    {
+      "agentId": "po",
+      "canonicalProfile": "agents/profiles/po.md",
+      "adapterPath": ".claude/commands/gos/agents/po.md"
+    },
+    {
+      "agentId": "qa",
+      "canonicalProfile": "agents/profiles/qa.md",
+      "adapterPath": ".claude/commands/gos/agents/qa.md"
+    },
+    {
+      "agentId": "devops",
+      "canonicalProfile": "agents/profiles/devops.md",
+      "adapterPath": ".claude/commands/gos/agents/devops.md"
+    },
+    {
+      "agentId": "squad-creator",
+      "canonicalProfile": "agents/profiles/squad-creator.md",
+      "adapterPath": ".claude/commands/gos/agents/squad-creator.md"
+    }
+  ]
+}

package/integrations/claude/command-map.json

@@ -0,0 +1,32 @@
+{
+  "ide": "claude",
+  "version": "1.0.0",
+  "lastUpdated": "2026-03-01",
+  "mappings": [
+    {
+      "ideCommand": "/gos-search",
+      "canonicalCommand": "gos.search",
+      "coreArtifact": "prompts/01-search.md"
+    },
+    {
+      "ideCommand": "/gos-spec",
+      "canonicalCommand": "gos.spec",
+      "coreArtifact": "prompts/02-spec.md"
+    },
+    {
+      "ideCommand": "/gos-tasks",
+      "canonicalCommand": "gos.tasks",
+      "coreArtifact": "prompts/03-tasks.md"
+    },
+    {
+      "ideCommand": "/gos-code",
+      "canonicalCommand": "gos.code",
+      "coreArtifact": "prompts/04-code.md"
+    },
+    {
+      "ideCommand": "/gos-review",
+      "canonicalCommand": "gos.review",
+      "coreArtifact": "prompts/05-reviews.md"
+    }
+  ]
+}

package/integrations/claude/litellm-proxy.md

@@ -0,0 +1,93 @@
+# LiteLLM Proxy — Modelos alternativos no Claude Code
+
+**Origem:** `docs/sources/integrations/litellm-openrouter.md`
+
+---
+
+## Conceito
+
+Usar o LiteLLM como proxy para rotear requests do Claude Code para modelos mais baratos ou gratuitos via OpenRouter.
+
+---
+
+## Setup
+
+### 1. Instalar LiteLLM
+
+```bash
+python3 -m venv ~/litellm-env
+source ~/litellm-env/bin/activate
+pip install "litellm[proxy]"
+```
+
+### 2. Configurar OpenRouter
+
+1. Criar conta em openrouter.ai
+2. Gerar API key: Settings > Keys
+3. Exportar:
+
+```bash
+export OPENROUTER_API_KEY=sk-or-v1-...
+```
+
+### 3. Rodar proxy
+
+```bash
+source ~/litellm-env/bin/activate
+export OPENROUTER_API_KEY=sk-or-v1-sua-chave
+litellm --model openrouter/z-ai/glm-5 --port 4000
+```
+
+### 4. Apontar Claude Code
+
+```bash
+export ANTHROPIC_BASE_URL=http://localhost:4000
+claude
+```
+
+---
+
+## Trocar modelo (reiniciar proxy)
+
+```bash
+# Modelo pago barato
+litellm --model openrouter/minimax/minimax-m2.5 --port 4000
+
+# Modelo gratuito
+litellm --model openrouter/z-ai/glm-4.5-air:free --port 4000
+
+# Modelo local via Ollama
+ollama pull deepseek-coder-v3
+litellm --model ollama/deepseek-coder-v3 --port 4000
+```
+
+---
+
+## Tabela de custos
+
+| Rota | Modelo | Input/M tokens |
+|------|--------|----------------|
+| Direto | claude-opus-4-6 | $15.00 |
+| Direto | claude-sonnet-4-5 | $3.00 |
+| OpenRouter | glm-5 | $0.75 |
+| OpenRouter | minimax-m2.5 | $0.20 |
+| OpenRouter | glm-4.5-air:free | $0.00 |
+| Local (Ollama) | deepseek-coder-v3 | $0.00 |
+
+---
+
+## Verificacao
+
+1. Fazer pergunta ao Claude Code → verificar resposta
+2. Visitar openrouter.ai/activity → ver logs com custo
+3. `/cost` no Claude Code mostra $0.00 (ja via proxy)
+
+---
+
+## Notas
+
+- Oficialmente permitido pela Anthropic (`ANTHROPIC_BASE_URL` e documentado)
+- Qualidade varia por modelo — Opus e premium por uma razao
+- LiteLLM deve ficar rodando enquanto Claude Code estiver ativo
+- Agent Teams funcionam (todos usam o mesmo proxy)
+- **NAO recomendado** para: code review critico, arquitetura complexa, seguranca

package/integrations/claude/mcp-specifics.md

@@ -0,0 +1,121 @@
+# MCP Configuration - Claude Code Specifics
+
+> Este documento complementa `rules/mcp-usage.md` (governança genérica MCP)
+> com configuração, seleção de ferramentas e troubleshooting específicos do Claude Code.
+
+## MCP Configuration Architecture
+
+G-OS usa **Docker MCP Toolkit** como infraestrutura principal para MCP.
+
+### Global (~/.claude.json) — MCP_DOCKER
+
+O gateway Docker MCP Toolkit é registrado globalmente no Claude Code:
+
+```json
+"MCP_DOCKER": {
+  "command": "docker",
+  "args": ["mcp", "gateway", "run"],
+  "env": {
+    "LOCALAPPDATA": "C:\\Users\\<user>\\AppData\\Local",
+    "ProgramData": "C:\\ProgramData",
+    "ProgramFiles": "C:\\Program Files"
+  }
+}
+```
+
+### Servers disponíveis via Docker MCP Toolkit
+
+| MCP | Tools | Purpose |
+|-----|-------|---------|
+| **Playwright** | 22 | Browser automation, screenshots, web testing |
+| **Context7** | 2 | Library documentation lookup (resolve-library-id + get-library-docs) |
+| **Chroma** | 13 | Vector database para RAG/embeddings |
+| **Sequential Thinking** | 1 | Raciocínio estruturado passo-a-passo |
+| **Fetch** | 1 | Web fetch com markdown extraction |
+| **MCP Gateway** | 5 | mcp-add, mcp-remove, mcp-find, mcp-exec, mcp-config-set |
+| **Code Mode** | 1 | Combinar múltiplas tools em JavaScript |
+
+**Total: 45 tools**
+
+### Outros MCPs globais (opcionais)
+
+| MCP | Purpose |
+|-----|---------|
+| **context7** (stdio) | Context7 direto via npx (alternativa ao Docker) |
+| **g-os** (http) | G-OS framework API |
+
+## CRITICAL: Tool Selection Priority
+
+SEMPRE preferir native Claude Code tools sobre MCP servers:
+
+| Task | USE THIS | NOT THIS |
+|------|----------|----------|
+| Read files | `Read` tool | MCP_DOCKER |
+| Write files | `Write` / `Edit` tools | MCP_DOCKER |
+| Run commands | `Bash` tool | MCP_DOCKER |
+| Search files | `Glob` tool | MCP_DOCKER |
+| Search content | `Grep` tool | MCP_DOCKER |
+| List directories | `Bash(ls)` or `Glob` | MCP_DOCKER |
+
+## Docker MCP Toolkit Usage
+
+### ONLY use MCP_DOCKER tools when:
+1. User explicitly asks for browser automation → **Playwright tools**
+2. Task requires vector search/RAG → **Chroma tools**
+3. Library documentation lookup → **Context7 tools**
+4. Web page fetching → **Fetch tool**
+5. Need to add/remove MCP servers dynamically → **MCP Gateway tools**
+6. Complex multi-step reasoning → **Sequential Thinking**
+
+### Managing servers via Docker MCP Toolkit
+
+```bash
+# List all tools available
+docker mcp tools ls
+
+# Inspect a specific tool
+docker mcp tools inspect <tool-name>
+
+# Test a tool
+docker mcp tools call <tool-name>
+
+# Find new servers in catalog (316+ available)
+docker mcp gateway catalog
+
+# Add a server from catalog
+# Use Docker Desktop UI → MCP Toolkit → Catalog tab
+```
+
+## Known Issues
+
+### Docker MCP Secrets Bug (Dec 2025)
+
+**Issue:** Docker MCP Toolkit's secrets store and template interpolation do not work properly. Credentials set via `docker mcp secret set` are NOT passed to containers.
+
+**Symptoms:**
+- `docker mcp tools ls` shows "(N prompts)" instead of "(N tools)"
+- MCP server starts but fails authentication
+- Verbose output shows `-e ENV_VAR` without values
+
+**Workaround:** Edit `~/.docker/mcp/catalogs/docker-mcp.yaml` directly with hardcoded env values:
+```yaml
+{mcp-name}:
+  env:
+    - name: API_TOKEN
+      value: 'actual-token-value'
+```
+
+**Affected MCPs:** Any MCP requiring authentication (Apify, Notion, Slack, etc.)
+
+### Docker not in PATH (Windows)
+
+Se `docker` não estiver no PATH do shell do Claude Code, o gateway pode falhar.
+
+**Fix:** Docker Desktop deve adicionar ao PATH automaticamente. Se não:
+```
+C:\Program Files\Docker\Docker\resources\bin
+```
+
+## Bootstrap para novos projetos
+
+Ver: `docs/sources/mcp/docker-mcp-toolkit-setup.md`

package/integrations/codex/README.md

@@ -0,0 +1,29 @@
+# Codex Adapter
+
+Adapter para ambiente Codex (OpenAI CLI).
+
+## Status: Ativo
+
+## Como funciona
+
+O Codex le automaticamente o arquivo `AGENTS.md` na raiz do projeto.
+O script `sync-codex-agents.js` gera esse arquivo a partir das fontes canonicas do framework.
+
+## Sync
+
+```bash
+npm run sync:codex-agents        # Gera/atualiza AGENTS.md
+npm run check:codex-agents       # Verifica se esta em sincronia
+```
+
+## Fontes
+
+- `agents/profiles/index.json` — perfis de agentes
+- `skills/registry.json` — skills registradas
+- `rules/*.md` — regras de governanca
+- `playbooks/*.md` — playbooks operacionais
+
+## Referencia
+
+- Command map: `integrations/codex/command-map.json`
+- Sync script: `scripts/integrations/sync-codex-agents.js`

package/integrations/codex/command-map.json

@@ -0,0 +1,29 @@
+{
+  "ide": "codex",
+  "version": "1.0.0",
+  "status": "active",
+  "mappings": [
+    {
+      "canonicalCommand": "gos.search",
+      "coreArtifact": "prompts/01-search.md"
+    },
+    {
+      "canonicalCommand": "gos.spec",
+      "coreArtifact": "prompts/02-spec.md"
+    },
+    {
+      "canonicalCommand": "gos.tasks",
+      "coreArtifact": "prompts/03-tasks.md"
+    },
+    {
+      "canonicalCommand": "gos.code",
+      "coreArtifact": "prompts/04-code.md"
+    },
+    {
+      "canonicalCommand": "gos.review",
+      "coreArtifact": "prompts/05-reviews.md"
+    }
+  ]
+}
+
+

package/integrations/cursor/README.md

@@ -0,0 +1,8 @@
+# Cursor Adapter
+
+Cursor no `G-OS` usa:
+
+- `AGENTS.md` como contexto compartilhado
+- `CLAUDE.md` como complemento
+- `.cursor/rules/g-os.mdc` como regra dedicada do projeto
+- `.cursorrules` como ponte de compatibilidade para setups antigos

package/integrations/cursor/command-map.json

@@ -0,0 +1,11 @@
+{
+  "ide": "cursor",
+  "version": "1.0.0",
+  "mappings": [
+    { "canonicalCommand": "gos.search", "coreArtifact": "prompts/01-search.md" },
+    { "canonicalCommand": "gos.spec", "coreArtifact": "prompts/02-spec.md" },
+    { "canonicalCommand": "gos.tasks", "coreArtifact": "prompts/03-tasks.md" },
+    { "canonicalCommand": "gos.code", "coreArtifact": "prompts/04-code.md" },
+    { "canonicalCommand": "gos.review", "coreArtifact": "prompts/05-reviews.md" }
+  ]
+}

package/integrations/gemini/README.md

@@ -0,0 +1,13 @@
+# Gemini CLI Adapter
+
+Gemini CLI funciona no `G-OS` com dois pontos principais:
+
+- `GEMINI.md` como arquivo de contexto de projeto
+- `.gemini/skills/gos-{slug}/SKILL.md` para descoberta de skills curadas
+
+## Setup recomendado
+
+- apontar o contexto do projeto para `GEMINI.md`
+- manter `AGENTS.md` e `CLAUDE.md` como referencias auxiliares
+- usar `.gemini/skills/` como espelho leve das skills do `G-OS`
+- regenerar adapters com `npm run sync:ides`

package/integrations/gemini/command-map.json

@@ -0,0 +1,11 @@
+{
+  "ide": "gemini-cli",
+  "version": "1.0.0",
+  "mappings": [
+    { "canonicalCommand": "gos.search", "coreArtifact": "prompts/01-search.md" },
+    { "canonicalCommand": "gos.spec", "coreArtifact": "prompts/02-spec.md" },
+    { "canonicalCommand": "gos.tasks", "coreArtifact": "prompts/03-tasks.md" },
+    { "canonicalCommand": "gos.code", "coreArtifact": "prompts/04-code.md" },
+    { "canonicalCommand": "gos.review", "coreArtifact": "prompts/05-reviews.md" }
+  ]
+}

package/integrations/kilo-code/README.md

@@ -0,0 +1,7 @@
+# Kilo Code Adapter
+
+Kilo Code no `G-OS` usa:
+
+- `AGENTS.md` e `CLAUDE.md`
+- `.kilocode/rules/g-os.md`
+- `.kilocode/workflows/`

package/integrations/kilo-code/command-map.json

@@ -0,0 +1,11 @@
+{
+  "ide": "kilo-code",
+  "version": "1.0.0",
+  "mappings": [
+    { "canonicalCommand": "gos.search", "coreArtifact": "prompts/01-search.md" },
+    { "canonicalCommand": "gos.spec", "coreArtifact": "prompts/02-spec.md" },
+    { "canonicalCommand": "gos.tasks", "coreArtifact": "prompts/03-tasks.md" },
+    { "canonicalCommand": "gos.code", "coreArtifact": "prompts/04-code.md" },
+    { "canonicalCommand": "gos.review", "coreArtifact": "prompts/05-reviews.md" }
+  ]
+}

package/integrations/opencode/README.md

@@ -0,0 +1,91 @@
+# OpenCode Adapter
+
+Adapter ativo para ambiente OpenCode (opencode.ai).
+
+## Status: Ativo
+
+- Skills sincronizadas automaticamente via `npm run sync:ides`
+- 73 skills do G-OS disponiveis como `.opencode/skills/gos-{slug}/SKILL.md`
+- `npm run sync:ides` gera adapters automaticamente
+
+## Compatibilidade nativa
+
+OpenCode le nativamente:
+- `AGENTS.md` (mesmo formato que Codex)
+- `CLAUDE.md` (compatibilidade retroativa)
+- `.claude/skills/` (skills do Claude Code)
+
+Isso significa que projetos com G-OS ja funcionam no OpenCode sem configuracao adicional.
+
+## Setup
+
+```bash
+# 1. Sincronizar adapters do G-OS no projeto
+npm run sync:ides
+
+# 2. OpenCode detecta automaticamente:
+#    - AGENTS.md (regras do projeto)
+#    - .opencode/skills/ (skills do G-OS)
+#    - CLAUDE.md (instrucoes adicionais)
+
+# 3. Usar skills no OpenCode
+# O agente descobre skills via tool "skill" e carrega sob demanda
+```
+
+## Sync manual
+
+```bash
+# Regenerar adapters
+npm run sync:ides
+
+# Validar adapters
+npm run check:ides
+
+# Regenerar e validar tudo
+npm run doctor
+```
+
+## Equivalencia G-OS <-> OpenCode
+
+| G-OS | OpenCode |
+|-----|----------|
+| `.G-OS/skills/*/SKILL.md` | `.opencode/skills/*/SKILL.md` |
+| `agents/.base/index.json` | `opencode.json` → `agent` |
+| `rules/*.md` | `AGENTS.md` ou `instructions[]` |
+| `playbooks/*.md` | `.opencode/commands/*.md` |
+| MCP via `.claude/settings.json` | MCP via `opencode.json` → `mcp` |
+
+## Configuracao recomendada
+
+```json
+{
+  "$schema": "https://opencode.ai/config.json",
+  "model": "anthropic/claude-sonnet-4-5",
+  "instructions": ["AGENTS.md", "CLAUDE.md"],
+  "permission": {
+    "skill": { "*": "allow" }
+  }
+}
+```
+
+## Arquitetura
+
+```
+projeto/
+├── AGENTS.md                          # OpenCode le nativamente
+├── CLAUDE.md                          # OpenCode le nativamente
+├── opencode.json                      # Config OpenCode (opcional)
+├── .opencode/
+│   └── skills/
+│       ├── gos-commit-dev/SKILL.md     # Adapter auto-gerado
+│       ├── gos-fe/SKILL.md             # Adapter auto-gerado
+│       └── ...                        # skills sincronizadas
+└── .G-OS/
+    └── skills/                        # Fonte canonica
+```
+
+## Referencia
+
+- [Documentacao OpenCode](https://opencode.ai/docs/)
+- `integrations/opencode/command-map.json` — mapeamento de comandos
+- `scripts/integrations/setup-ide-adapters.js` — script de sync

package/integrations/opencode/command-map.json

@@ -0,0 +1,46 @@
+{
+  "ide": "opencode",
+  "version": "2.0.0",
+  "status": "active",
+  "notes": {
+    "compatReads": "OpenCode le AGENTS.md e CLAUDE.md nativamente, alem de .claude/skills/",
+    "skillsSync": "npm run sync:ides gera .opencode/skills/gos-{slug}/SKILL.md",
+    "configFile": "opencode.json (projeto) ou ~/.config/opencode/opencode.json (global)"
+  },
+  "mappings": [
+    {
+      "canonicalCommand": "gos.search",
+      "coreArtifact": "prompts/01-search.md",
+      "opencodeEquiv": "commands/search.md ou skill gos-sdd"
+    },
+    {
+      "canonicalCommand": "gos.spec",
+      "coreArtifact": "prompts/02-spec.md",
+      "opencodeEquiv": "commands/spec.md ou skill gos-sdd"
+    },
+    {
+      "canonicalCommand": "gos.tasks",
+      "coreArtifact": "prompts/03-tasks.md",
+      "opencodeEquiv": "commands/tasks.md ou skill gos-plan-to-tasks"
+    },
+    {
+      "canonicalCommand": "gos.code",
+      "coreArtifact": "prompts/04-code.md",
+      "opencodeEquiv": "commands/code.md ou skill gos-be / gos-fe"
+    },
+    {
+      "canonicalCommand": "gos.review",
+      "coreArtifact": "prompts/05-reviews.md",
+      "opencodeEquiv": "commands/review.md ou skill gos-review-pr"
+    }
+  ],
+  "equivalenceTable": {
+    "gos → opencode": {
+      ".G-OS/skills/*/SKILL.md": ".opencode/skills/gos-*/SKILL.md",
+      "agents/.base/index.json": "opencode.json → agent {}",
+      "rules/*.md": "AGENTS.md ou instructions[]",
+      "playbooks/*.md": ".opencode/commands/*.md",
+      "MCP (.claude/settings.json)": "opencode.json → mcp {}"
+    }
+  }
+}

package/integrations/registry.json

@@ -0,0 +1,20 @@
+{
+  "version": "0.1.0",
+  "lastUpdated": "2026-03-30",
+  "canonicalCommands": [
+    { "id": "gos.search", "coreArtifact": "prompts/01-search.md" },
+    { "id": "gos.spec", "coreArtifact": "prompts/02-spec.md" },
+    { "id": "gos.tasks", "coreArtifact": "prompts/03-tasks.md" },
+    { "id": "gos.code", "coreArtifact": "prompts/04-code.md" },
+    { "id": "gos.review", "coreArtifact": "prompts/05-reviews.md" }
+  ],
+  "supportedIdes": {
+    "claude-code": { "workspaceDir": ".claude", "adapter": "integrations/claude", "status": "active" },
+    "antigravity": { "workspaceDir": ".antigravity", "adapter": "integrations/antigravity", "status": "active" },
+    "gemini-cli": { "workspaceDir": ".gemini", "contextFile": "GEMINI.md", "status": "active" },
+    "cursor": { "workspaceDir": ".cursor", "ruleFile": ".cursor/rules/g-os.mdc", "status": "active" },
+    "codex": { "workspaceDir": ".codex", "adapter": "integrations/codex", "status": "active" },
+    "opencode": { "workspaceDir": ".opencode", "adapter": "integrations/opencode", "status": "active" },
+    "kilo-code": { "workspaceDir": ".kilocode", "ruleFile": ".kilocode/rules/g-os.md", "status": "active" }
+  }
+}

package/manifests/g-os-runtime-manifest.json

@@ -0,0 +1,39 @@
+{
+  "runtime": "ganbatte-os-design-delivery",
+  "version": "0.1.0",
+  "description": "Runtime curado do ganbatte-os para design-to-code, sprint planning e ClickUp delivery.",
+  "modules": {
+    "agents": [
+      "ganbatte-os-master",
+      "ux-design-expert",
+      "architect",
+      "dev",
+      "sm",
+      "po",
+      "devops",
+      "squad-creator"
+    ],
+    "skills": [
+      "design-to-code",
+      "figma-implement-design",
+      "figma-make-analyzer",
+      "make-code-triage",
+      "make-version-diff",
+      "component-dedup",
+      "frontend-dev",
+      "react-best-practices",
+      "react-doctor",
+      "sprint-planner",
+      "clickup",
+      "plan-to-tasks",
+      "agent-teams",
+      "git-ssh-setup"
+    ],
+    "squads": [
+      "design-delivery",
+      "design-squad",
+      "sprint-planning",
+      "git-operations"
+    ]
+  }
+}