@dewtech/dare-cli 3.3.0 → 3.5.0

package/dist/utils/templates.js

@@ -1,473 +1,473 @@
 export function generateCursorRules(config) {
     const { backend, frontend, graphrag, mcp } = config;
     const backendRules = {
-        'rust-axum': `## Backend: Rust/Axum
-- Use Rust idioms and patterns (no unwrap in production)
-- Prefer async/await with Tokio runtime
-- Use Axum extractors for request handling
-- Handle all errors with thiserror/anyhow
-- Run \`cargo clippy\` and \`cargo test\` before completing any task (Ralph Loop)
+        'rust-axum': `## Backend: Rust/Axum
+- Use Rust idioms and patterns (no unwrap in production)
+- Prefer async/await with Tokio runtime
+- Use Axum extractors for request handling
+- Handle all errors with thiserror/anyhow
+- Run \`cargo clippy\` and \`cargo test\` before completing any task (Ralph Loop)
 - Use SQLx for database queries with compile-time verification`,
-        'node-nestjs': `## Backend: Node.js/NestJS
-- Use NestJS decorators and dependency injection
-- Define DTOs with class-validator for all inputs
-- Use TypeORM or Prisma for database access
-- Write Jest unit and e2e tests for all services
+        'node-nestjs': `## Backend: Node.js/NestJS
+- Use NestJS decorators and dependency injection
+- Define DTOs with class-validator for all inputs
+- Use TypeORM or Prisma for database access
+- Write Jest unit and e2e tests for all services
 - Follow NestJS module structure`,
-        'python-fastapi': `## Backend: Python/FastAPI
-- Use Pydantic v2 for all data validation
-- Type all functions with PEP 484 type hints
-- Use async/await for all IO operations
-- Follow PEP 8 and use ruff for linting
+        'python-fastapi': `## Backend: Python/FastAPI
+- Use Pydantic v2 for all data validation
+- Type all functions with PEP 484 type hints
+- Use async/await for all IO operations
+- Follow PEP 8 and use ruff for linting
 - Write pytest tests with coverage`,
-        'php-laravel': `## Backend: PHP/Laravel
-- Follow PSR-12 coding standards
-- Use FormRequests for all input validation
-- Use API Resources for all responses
-- Write PHPUnit feature and unit tests
+        'php-laravel': `## Backend: PHP/Laravel
+- Follow PSR-12 coding standards
+- Use FormRequests for all input validation
+- Use API Resources for all responses
+- Write PHPUnit feature and unit tests
 - Use Eloquent ORM with proper relationships`,
     };
     const frontendRules = {
-        react: `## Frontend: React 18+
-- Use functional components with hooks only
-- TypeScript for all components and hooks
-- Prefer React Query for server state management
-- Use Zustand or Context API for client state
+        react: `## Frontend: React 18+
+- Use functional components with hooks only
+- TypeScript for all components and hooks
+- Prefer React Query for server state management
+- Use Zustand or Context API for client state
 - Write Vitest + Testing Library tests`,
-        vue: `## Frontend: Vue 3+
-- Use Composition API with <script setup> syntax
-- TypeScript for all components
-- Use Pinia for state management
-- Use Vue Router for navigation
+        vue: `## Frontend: Vue 3+
+- Use Composition API with <script setup> syntax
+- TypeScript for all components
+- Use Pinia for state management
+- Use Vue Router for navigation
 - Write Vitest + Vue Test Utils tests`,
-        'rust-leptos': `## Frontend: Leptos fullstack (Rust SSR + WASM)
-- Use \`#[component]\` macro — no class components
-- State: \`signal()\`, \`Resource\` for async, \`Action\` for mutations
-- Loading: \`Suspense\`/\`Transition\` — never block the render
-- Server functions: \`#[server]\` macro (ssr feature only)
-- Shared types: \`#[cfg_attr(feature = "ssr", derive(sqlx::FromRow))]\`
-- Avoid: \`tokio::spawn\` on client, \`panic!\` in components, \`Effect\` for fetch
+        'rust-leptos': `## Frontend: Leptos fullstack (Rust SSR + WASM)
+- Use \`#[component]\` macro — no class components
+- State: \`signal()\`, \`Resource\` for async, \`Action\` for mutations
+- Loading: \`Suspense\`/\`Transition\` — never block the render
+- Server functions: \`#[server]\` macro (ssr feature only)
+- Shared types: \`#[cfg_attr(feature = "ssr", derive(sqlx::FromRow))]\`
+- Avoid: \`tokio::spawn\` on client, \`panic!\` in components, \`Effect\` for fetch
 - Build: \`cargo leptos build\` | Test: \`cargo test --workspace\` | Dev: \`cargo leptos watch\``,
-        'rust-leptos-csr': `## Frontend: Leptos CSR (Rust WASM + trunk)
-- Pure client-side WASM — no SSR, no \`#[server]\` functions
-- Use \`#[component]\` macro, \`signal()\` for state, \`Resource\` for async
-- Build tool: \`trunk build\` (NOT cargo leptos) | Dev: \`trunk serve\`
+        'rust-leptos-csr': `## Frontend: Leptos CSR (Rust WASM + trunk)
+- Pure client-side WASM — no SSR, no \`#[server]\` functions
+- Use \`#[component]\` macro, \`signal()\` for state, \`Resource\` for async
+- Build tool: \`trunk build\` (NOT cargo leptos) | Dev: \`trunk serve\`
 - Test: \`cargo test --workspace\` | Deploy: \`dist/\` is fully static`,
     };
-    return `# DARE Framework - Cursor Rules
-
-## DARE Methodology
-You are an AI assistant following the DARE methodology:
-- **D**esign: Define requirements and architecture
-- **A**rchitect: Create technical blueprint and task graph
-- **R**eview: Validate implementation against blueprint
-- **E**xecute: Implement tasks following the DAG
-
-## Core Rules
-- Always read DARE/BLUEPRINT.md before implementing any feature
-- Update DARE/TASKS.md status after completing each task
-- Never skip the Ralph Loop (build → test → lint) before marking a task as DONE
-- Human approval is required before merging to main branch
-- Context is king: use MCP Server queries instead of re-reading large files
-
-## Project Structure
-- DARE/ - Methodology files (DESIGN.md, BLUEPRINT.md, TASKS.md, dare-dag.yaml)
-- DARE/EXECUTION/ - Task execution logs
-
-${backend && backendRules[backend] ? backendRules[backend] : ''}
-
-${frontend && frontendRules[frontend] ? frontendRules[frontend] : ''}
-
-## GraphRAG Context (${graphrag})
-${mcp ? `- Query MCP Server at http://localhost:3000 for context instead of reading full files
-- Use POST /context/query with {"type": "file"|"task"|"dependency", "query": "..."}` : '- Use DARE/BLUEPRINT.md as the single source of truth for context'}
-
-## Ralph Loop (Mandatory before DONE)
-1. Build the project (cargo build / npm run build / etc)
-2. Run tests (cargo test / npm test / pytest / etc)
-3. Run linter (cargo clippy / eslint / ruff / phpstan)
-4. Only mark task as DONE if all 3 steps pass
+    return `# DARE Framework - Cursor Rules
+
+## DARE Methodology
+You are an AI assistant following the DARE methodology:
+- **D**esign: Define requirements and architecture
+- **A**rchitect: Create technical blueprint and task graph
+- **R**eview: Validate implementation against blueprint
+- **E**xecute: Implement tasks following the DAG
+
+## Core Rules
+- Always read DARE/BLUEPRINT.md before implementing any feature
+- Update DARE/TASKS.md status after completing each task
+- Never skip the Ralph Loop (build → test → lint) before marking a task as DONE
+- Human approval is required before merging to main branch
+- Context is king: use MCP Server queries instead of re-reading large files
+
+## Project Structure
+- DARE/ - Methodology files (DESIGN.md, BLUEPRINT.md, TASKS.md, dare-dag.yaml)
+- DARE/EXECUTION/ - Task execution logs
+
+${backend && backendRules[backend] ? backendRules[backend] : ''}
+
+${frontend && frontendRules[frontend] ? frontendRules[frontend] : ''}
+
+## GraphRAG Context (${graphrag})
+${mcp ? `- Query MCP Server at http://localhost:3000 for context instead of reading full files
+- Use POST /context/query with {"type": "file"|"task"|"dependency", "query": "..."}` : '- Use DARE/BLUEPRINT.md as the single source of truth for context'}
+
+## Ralph Loop (Mandatory before DONE)
+1. Build the project (cargo build / npm run build / etc)
+2. Run tests (cargo test / npm test / pytest / etc)
+3. Run linter (cargo clippy / eslint / ruff / phpstan)
+4. Only mark task as DONE if all 3 steps pass
 `;
 }
 export function generateAntigravityRules(config) {
     const { backend, frontend, graphrag, mcp } = config;
-    return `# DARE Framework - Antigravity Rules
-
-## Agent Configuration
-You are an autonomous AI agent following the DARE methodology.
-Execute tasks from DARE/dare-dag.yaml in parallel when dependencies allow.
-
-## DARE Phases
-- **Design**: Read DARE/DESIGN.md for requirements
-- **Architect**: Read DARE/BLUEPRINT.md for technical spec
-- **Review**: Validate against blueprint before marking DONE
-- **Execute**: Implement tasks, update DARE/TASKS.md
-
-## Stack
-${backend ? `- Backend: ${backend}` : ''}
-${frontend ? `- Frontend: ${frontend}` : ''}
-
-## Context Strategy (${graphrag})
-${mcp ? `- Query MCP Server at http://localhost:3000 for context
-- Avoid reading full files; use targeted queries` : '- Read DARE/BLUEPRINT.md for context'}
-
-## Execution Rules
-- Always check task dependencies before starting
-- Update task status in DARE/TASKS.md in real-time
-- Run Ralph Loop before marking any task as DONE
-- Request human review for architectural decisions
+    return `# DARE Framework - Antigravity Rules
+
+## Agent Configuration
+You are an autonomous AI agent following the DARE methodology.
+Execute tasks from DARE/dare-dag.yaml in parallel when dependencies allow.
+
+## DARE Phases
+- **Design**: Read DARE/DESIGN.md for requirements
+- **Architect**: Read DARE/BLUEPRINT.md for technical spec
+- **Review**: Validate against blueprint before marking DONE
+- **Execute**: Implement tasks, update DARE/TASKS.md
+
+## Stack
+${backend ? `- Backend: ${backend}` : ''}
+${frontend ? `- Frontend: ${frontend}` : ''}
+
+## Context Strategy (${graphrag})
+${mcp ? `- Query MCP Server at http://localhost:3000 for context
+- Avoid reading full files; use targeted queries` : '- Read DARE/BLUEPRINT.md for context'}
+
+## Execution Rules
+- Always check task dependencies before starting
+- Update task status in DARE/TASKS.md in real-time
+- Run Ralph Loop before marking any task as DONE
+- Request human review for architectural decisions
 `;
 }
 export function generateMcpCursorRules(config) {
     const { mcpTransport = 'stdio', mcpLanguage = 'node-ts', mcpFeatures = ['tools'], graphrag, mcp } = config;
     const featuresStr = mcpFeatures.join(', ');
     const langRules = mcpLanguage === 'python'
-        ? `## Language: Python
-- Use FastMCP (@mcp.tool, @mcp.resource, @mcp.prompt decorators)
-- Type all arguments — FastMCP builds JSON schema from type hints
-- Use docstrings as tool/resource/prompt descriptions
-- Test: npx @modelcontextprotocol/inspector python main.py
+        ? `## Language: Python
+- Use FastMCP (@mcp.tool, @mcp.resource, @mcp.prompt decorators)
+- Type all arguments — FastMCP builds JSON schema from type hints
+- Use docstrings as tool/resource/prompt descriptions
+- Test: npx @modelcontextprotocol/inspector python main.py
 - Lint: ruff check . && mypy .`
-        : `## Language: TypeScript
-- Import from '@modelcontextprotocol/sdk/server/index.js' and types
-- Always define strict inputSchema for every tool (Claude depends on it)
-- Use zod for runtime validation when needed
-- Test: npm run inspect (MCP Inspector)
+        : `## Language: TypeScript
+- Import from '@modelcontextprotocol/sdk/server/index.js' and types
+- Always define strict inputSchema for every tool (Claude depends on it)
+- Use zod for runtime validation when needed
+- Test: npm run inspect (MCP Inspector)
 - Build: npm run build before marking any task DONE`;
     const transportNote = mcpTransport === 'stdio'
         ? `- Transport: stdio — server communicates via stdin/stdout, no HTTP port needed`
         : mcpTransport === 'sse'
             ? `- Transport: SSE — server exposes GET /sse and POST /messages endpoints`
             : `- Transport: HTTP Stream — use StreamableHTTPServerTransport`;
-    return `# DARE Framework - Cursor Rules (MCP Server Project)
-
-## DARE Methodology
-You are an AI assistant following the DARE methodology:
-- **D**esign: Define MCP server requirements and capabilities
-- **A**rchitect: Create technical blueprint and task graph
-- **R**eview: Validate implementation against blueprint
-- **E**xecute: Implement tools, resources, and prompts following the DAG
-
-## Core Rules
-- Always read DARE/BLUEPRINT.md before implementing any tool or resource
-- Update DARE/TASKS.md status after completing each task
-- Never skip the Ralph Loop (build → test → inspect) before marking a task as DONE
-- Test every tool with MCP Inspector before marking DONE
-- Validate tool inputSchema matches actual handler logic exactly
-
-## MCP Server Configuration
-- Transport: ${mcpTransport}
-- Features: ${featuresStr}
-${transportNote}
-
-${langRules}
-
-## MCP Best Practices
-- Keep tool names snake_case and descriptive
-- Return structured content arrays, not plain strings
-- Handle unknown tool/resource/prompt names with explicit errors
-- Never expose secrets via tool outputs or resource contents
-- Document every tool argument in inputSchema description fields
-
-## GraphRAG Context (${graphrag})
-${mcp ? `- Query MCP Server at http://localhost:3000 for context instead of reading full files` : '- Use DARE/BLUEPRINT.md as the single source of truth'}
-
-## Ralph Loop (Mandatory before DONE)
-1. Build (npm run build / python -m py_compile)
-2. Test (npm test / pytest)
-3. Inspect with MCP Inspector to verify tool contracts
-4. Only mark DONE if all 3 steps pass
+    return `# DARE Framework - Cursor Rules (MCP Server Project)
+
+## DARE Methodology
+You are an AI assistant following the DARE methodology:
+- **D**esign: Define MCP server requirements and capabilities
+- **A**rchitect: Create technical blueprint and task graph
+- **R**eview: Validate implementation against blueprint
+- **E**xecute: Implement tools, resources, and prompts following the DAG
+
+## Core Rules
+- Always read DARE/BLUEPRINT.md before implementing any tool or resource
+- Update DARE/TASKS.md status after completing each task
+- Never skip the Ralph Loop (build → test → inspect) before marking a task as DONE
+- Test every tool with MCP Inspector before marking DONE
+- Validate tool inputSchema matches actual handler logic exactly
+
+## MCP Server Configuration
+- Transport: ${mcpTransport}
+- Features: ${featuresStr}
+${transportNote}
+
+${langRules}
+
+## MCP Best Practices
+- Keep tool names snake_case and descriptive
+- Return structured content arrays, not plain strings
+- Handle unknown tool/resource/prompt names with explicit errors
+- Never expose secrets via tool outputs or resource contents
+- Document every tool argument in inputSchema description fields
+
+## GraphRAG Context (${graphrag})
+${mcp ? `- Query MCP Server at http://localhost:3000 for context instead of reading full files` : '- Use DARE/BLUEPRINT.md as the single source of truth'}
+
+## Ralph Loop (Mandatory before DONE)
+1. Build (npm run build / python -m py_compile)
+2. Test (npm test / pytest)
+3. Inspect with MCP Inspector to verify tool contracts
+4. Only mark DONE if all 3 steps pass
 `;
 }
 export function generateMcpAntigravityRules(config) {
     const { mcpTransport = 'stdio', mcpLanguage = 'node-ts', mcpFeatures = ['tools'], graphrag, mcp } = config;
-    return `# DARE Framework - Antigravity Rules (MCP Server Project)
-
-## Agent Configuration
-You are an autonomous AI agent implementing an MCP server using the DARE methodology.
-Execute tasks from DARE/dare-dag.yaml in parallel when dependencies allow.
-
-## DARE Phases
-- **Design**: Read DARE/DESIGN.md — what tools/resources/prompts does this MCP server expose?
-- **Architect**: Read DARE/BLUEPRINT.md — tool schemas, transport, auth strategy
-- **Review**: Test each tool with MCP Inspector before marking DONE
-- **Execute**: Implement tasks, update DARE/TASKS.md
-
-## MCP Stack
-- Language: ${mcpLanguage}
-- Transport: ${mcpTransport}
-- Features: ${mcpFeatures.join(', ')}
-
-## Implementation Rules
-- Each tool must have a strict inputSchema — Claude uses it to call the tool
-- Test with MCP Inspector after implementing each tool
-- Never skip error handling for unknown tool names
-
-## Context Strategy (${graphrag})
-${mcp ? `- Query MCP Server at http://localhost:3000 for project context` : '- Read DARE/BLUEPRINT.md for context'}
-
-## Execution Rules
-- Always check task dependencies before starting
-- Update task status in DARE/TASKS.md in real-time
-- Run Ralph Loop before marking any task as DONE
-- Request human review for transport or auth design decisions
+    return `# DARE Framework - Antigravity Rules (MCP Server Project)
+
+## Agent Configuration
+You are an autonomous AI agent implementing an MCP server using the DARE methodology.
+Execute tasks from DARE/dare-dag.yaml in parallel when dependencies allow.
+
+## DARE Phases
+- **Design**: Read DARE/DESIGN.md — what tools/resources/prompts does this MCP server expose?
+- **Architect**: Read DARE/BLUEPRINT.md — tool schemas, transport, auth strategy
+- **Review**: Test each tool with MCP Inspector before marking DONE
+- **Execute**: Implement tasks, update DARE/TASKS.md
+
+## MCP Stack
+- Language: ${mcpLanguage}
+- Transport: ${mcpTransport}
+- Features: ${mcpFeatures.join(', ')}
+
+## Implementation Rules
+- Each tool must have a strict inputSchema — Claude uses it to call the tool
+- Test with MCP Inspector after implementing each tool
+- Never skip error handling for unknown tool names
+
+## Context Strategy (${graphrag})
+${mcp ? `- Query MCP Server at http://localhost:3000 for project context` : '- Read DARE/BLUEPRINT.md for context'}
+
+## Execution Rules
+- Always check task dependencies before starting
+- Update task status in DARE/TASKS.md in real-time
+- Run Ralph Loop before marking any task as DONE
+- Request human review for transport or auth design decisions
 `;
 }
 export function generateClaudeCodeRules(config) {
     const { backend, frontend, graphrag, mcp } = config;
     const backendRules = {
-        'rust-axum': `## Backend: Rust/Axum
-- Use Rust idioms — no \`unwrap()\` in production code
-- Prefer async/await with Tokio runtime
-- Use Axum extractors for request handling
-- Handle all errors with \`thiserror\`/\`anyhow\`
+        'rust-axum': `## Backend: Rust/Axum
+- Use Rust idioms — no \`unwrap()\` in production code
+- Prefer async/await with Tokio runtime
+- Use Axum extractors for request handling
+- Handle all errors with \`thiserror\`/\`anyhow\`
 - Ralph Loop: \`cargo clippy && cargo test\``,
-        'node-nestjs': `## Backend: Node.js/NestJS
-- Use NestJS decorators and dependency injection
-- Define DTOs with class-validator for all inputs
-- Use TypeORM or Prisma for database access
+        'node-nestjs': `## Backend: Node.js/NestJS
+- Use NestJS decorators and dependency injection
+- Define DTOs with class-validator for all inputs
+- Use TypeORM or Prisma for database access
 - Ralph Loop: \`npm run build && npm test && npx eslint src\``,
-        'python-fastapi': `## Backend: Python/FastAPI
-- Use Pydantic v2 for all data validation
-- Type all functions with PEP 484 type hints
-- Use async/await for all IO operations
+        'python-fastapi': `## Backend: Python/FastAPI
+- Use Pydantic v2 for all data validation
+- Type all functions with PEP 484 type hints
+- Use async/await for all IO operations
 - Ralph Loop: \`python -m pytest && ruff check .\``,
-        'php-laravel': `## Backend: PHP/Laravel
-- Follow PSR-12 coding standards
-- Use FormRequests for all input validation
-- Use API Resources for all responses
+        'php-laravel': `## Backend: PHP/Laravel
+- Follow PSR-12 coding standards
+- Use FormRequests for all input validation
+- Use API Resources for all responses
 - Ralph Loop: \`php artisan test && ./vendor/bin/phpstan analyse\``,
     };
     const frontendRules = {
-        react: `## Frontend: React 18+
-- Use functional components with hooks only
-- TypeScript for all components and hooks
-- Prefer React Query for server state management
+        react: `## Frontend: React 18+
+- Use functional components with hooks only
+- TypeScript for all components and hooks
+- Prefer React Query for server state management
 - Ralph Loop: \`npm run build && npm test && npx eslint src\``,
-        vue: `## Frontend: Vue 3+
-- Use Composition API with <script setup> syntax
-- TypeScript for all components
-- Use Pinia for state management
+        vue: `## Frontend: Vue 3+
+- Use Composition API with <script setup> syntax
+- TypeScript for all components
+- Use Pinia for state management
 - Ralph Loop: \`npm run build && npm test && npx eslint src\``,
-        'rust-leptos': `## Frontend: Leptos fullstack (Rust SSR + WASM)
-- Use \`#[component]\` macro — estado com \`signal()\`, async com \`Resource\`, mutações com \`Action\`
-- Loading: \`Suspense\`/\`Transition\` — nunca bloqueie o render
-- Server functions: \`#[server]\` macro (feature ssr)
-- Tipos compartilhados server + WASM: \`#[cfg_attr(feature = "ssr", derive(sqlx::FromRow))]\`
-- Evite: \`tokio::spawn\` no client, \`panic!\` em componentes, \`Effect\` para fetch
+        'rust-leptos': `## Frontend: Leptos fullstack (Rust SSR + WASM)
+- Use \`#[component]\` macro — estado com \`signal()\`, async com \`Resource\`, mutações com \`Action\`
+- Loading: \`Suspense\`/\`Transition\` — nunca bloqueie o render
+- Server functions: \`#[server]\` macro (feature ssr)
+- Tipos compartilhados server + WASM: \`#[cfg_attr(feature = "ssr", derive(sqlx::FromRow))]\`
+- Evite: \`tokio::spawn\` no client, \`panic!\` em componentes, \`Effect\` para fetch
 - Ralph Loop: \`cargo leptos build --release && cargo test --workspace && cargo clippy --all-features -- -D warnings\``,
-        'rust-leptos-csr': `## Frontend: Leptos CSR (Rust WASM + trunk)
-- WASM puro — sem SSR, sem \`#[server]\`
-- Build: \`trunk build --release\` | Dev: \`trunk serve\` | Test: \`cargo test --workspace\`
+        'rust-leptos-csr': `## Frontend: Leptos CSR (Rust WASM + trunk)
+- WASM puro — sem SSR, sem \`#[server]\`
+- Build: \`trunk build --release\` | Dev: \`trunk serve\` | Test: \`cargo test --workspace\`
 - Ralph Loop: \`trunk build --release && cargo test --workspace && cargo clippy --all-features -- -D warnings\``,
     };
-    return `# DARE Framework
-
-## Metodologia
-Você é o Claude Code, assistente de desenvolvimento seguindo o método DARE:
-- **D**esign: Requisitos e objetivos definidos em \`DARE/DESIGN.md\`
-- **A**rchitect: Blueprint técnico e grafo de tasks em \`DARE/BLUEPRINT.md\`
-- **R**eview: Validação humana antes de executar
-- **E**xecute: Implementação task a task com Ralph Loop
-
-## Regras Fundamentais
-- Sempre leia \`DARE/BLUEPRINT.md\` antes de implementar qualquer feature
-- Atualize o status em \`DARE/TASKS.md\` ao concluir cada task
-- Nunca pule o Ralph Loop (build → test → lint) antes de marcar uma task como DONE
-- Aprovação humana obrigatória antes de merge para a branch principal
-- Use os slash commands \`/dare-design\`, \`/dare-blueprint\`, \`/dare-execute\`
-
-## Estrutura do Projeto
-\`\`\`
-DARE/
-├── DESIGN.md        ← Fase D — requisitos (humano define)
-├── BLUEPRINT.md     ← Fase A — arquitetura (IA propõe, humano valida)
-├── TASKS.md         ← rastreamento de tasks
-├── dare-dag.yaml    ← grafo de dependências
-└── EXECUTION/       ← logs de execução por task
-\`\`\`
-
-${backend && backendRules[backend] ? backendRules[backend] : ''}
-
-${frontend && frontendRules[frontend] ? frontendRules[frontend] : ''}
-
-## Contexto (${graphrag})
-${mcp ? `- Consulte o DARE MCP Server em http://localhost:3000 para queries de contexto
-- Use POST /context/query com {"type": "architecture"|"task"|"dependency", "query": "..."}
-- Evite reler arquivos inteiros — use queries direcionadas` : `- Use DARE/BLUEPRINT.md como fonte única de verdade para contexto`}
-
-## Ralph Loop (obrigatório antes de DONE)
-1. Build — compile e verifique erros
-2. Test — rode a suite de testes completa
-3. Lint — rode o linter/formatter
-4. Só marque DONE se os 3 passos passarem sem erros
+    return `# DARE Framework
+
+## Metodologia
+Você é o Claude Code, assistente de desenvolvimento seguindo o método DARE:
+- **D**esign: Requisitos e objetivos definidos em \`DARE/DESIGN.md\`
+- **A**rchitect: Blueprint técnico e grafo de tasks em \`DARE/BLUEPRINT.md\`
+- **R**eview: Validação humana antes de executar
+- **E**xecute: Implementação task a task com Ralph Loop
+
+## Regras Fundamentais
+- Sempre leia \`DARE/BLUEPRINT.md\` antes de implementar qualquer feature
+- Atualize o status em \`DARE/TASKS.md\` ao concluir cada task
+- Nunca pule o Ralph Loop (build → test → lint) antes de marcar uma task como DONE
+- Aprovação humana obrigatória antes de merge para a branch principal
+- Use os slash commands \`/dare-design\`, \`/dare-blueprint\`, \`/dare-execute\`
+
+## Estrutura do Projeto
+\`\`\`
+DARE/
+├── DESIGN.md        ← Fase D — requisitos (humano define)
+├── BLUEPRINT.md     ← Fase A — arquitetura (IA propõe, humano valida)
+├── TASKS.md         ← rastreamento de tasks
+├── dare-dag.yaml    ← grafo de dependências
+└── EXECUTION/       ← logs de execução por task
+\`\`\`
+
+${backend && backendRules[backend] ? backendRules[backend] : ''}
+
+${frontend && frontendRules[frontend] ? frontendRules[frontend] : ''}
+
+## Contexto (${graphrag})
+${mcp ? `- Consulte o DARE MCP Server em http://localhost:3000 para queries de contexto
+- Use POST /context/query com {"type": "architecture"|"task"|"dependency", "query": "..."}
+- Evite reler arquivos inteiros — use queries direcionadas` : `- Use DARE/BLUEPRINT.md como fonte única de verdade para contexto`}
+
+## Ralph Loop (obrigatório antes de DONE)
+1. Build — compile e verifique erros
+2. Test — rode a suite de testes completa
+3. Lint — rode o linter/formatter
+4. Só marque DONE se os 3 passos passarem sem erros
 `;
 }
 export function generateMcpClaudeCodeRules(config) {
     const { mcpTransport = 'stdio', mcpLanguage = 'node-ts', mcpFeatures = ['tools'], graphrag, mcp } = config;
     const featuresStr = mcpFeatures.join(', ');
     const langRules = mcpLanguage === 'python'
-        ? `## Linguagem: Python
-- Use FastMCP com decoradores @mcp.tool, @mcp.resource, @mcp.prompt
-- Type hints obrigatórios — FastMCP gera o JSON Schema automaticamente
-- Use docstrings como descrição das tools/resources/prompts
-- Ralph Loop: \`python -m py_compile main.py && pytest && ruff check .\`
+        ? `## Linguagem: Python
+- Use FastMCP com decoradores @mcp.tool, @mcp.resource, @mcp.prompt
+- Type hints obrigatórios — FastMCP gera o JSON Schema automaticamente
+- Use docstrings como descrição das tools/resources/prompts
+- Ralph Loop: \`python -m py_compile main.py && pytest && ruff check .\`
 - Inspecione com: \`npx @modelcontextprotocol/inspector python main.py\``
-        : `## Linguagem: TypeScript
-- Importe de \`@modelcontextprotocol/sdk/server/index.js\`
-- Defina \`inputSchema\` estrito para cada tool — Claude depende dele para chamadas corretas
-- Use zod para validação em runtime quando necessário
-- Ralph Loop: \`npm run build && npm test\`
+        : `## Linguagem: TypeScript
+- Importe de \`@modelcontextprotocol/sdk/server/index.js\`
+- Defina \`inputSchema\` estrito para cada tool — Claude depende dele para chamadas corretas
+- Use zod para validação em runtime quando necessário
+- Ralph Loop: \`npm run build && npm test\`
 - Inspecione com: \`npm run inspect\``;
     const transportNote = mcpTransport === 'stdio'
         ? `- Transport \`stdio\` — comunicação via stdin/stdout, sem porta HTTP`
         : mcpTransport === 'sse'
             ? `- Transport \`SSE\` — expõe GET /sse e POST /messages`
             : `- Transport \`HTTP Stream\` — use StreamableHTTPServerTransport`;
-    return `# DARE Framework — MCP Server
-
-## Metodologia
-Você é o Claude Code implementando um servidor MCP com o método DARE:
-- **D**esign: Quais tools/resources/prompts este servidor expõe?
-- **A**rchitect: Schemas das tools, estratégia de transport e auth
-- **R**eview: Teste cada tool com MCP Inspector antes de marcar DONE
-- **E**xecute: Implementação task a task com Ralph Loop
-
-## Configuração do MCP Server
-- Transport: ${mcpTransport}
-- Features: ${featuresStr}
-${transportNote}
-
-## Regras Fundamentais
-- Leia \`DARE/BLUEPRINT.md\` antes de implementar qualquer tool
-- Atualize \`DARE/TASKS.md\` ao concluir cada task
-- Nunca pule o Ralph Loop antes de marcar DONE
-- Teste cada tool com MCP Inspector antes de marcar DONE
-- Valide que o \`inputSchema\` corresponde exatamente ao handler
-
-${langRules}
-
-## Boas Práticas MCP
-- Nomes de tools em snake_case e descritivos
-- Retorne arrays de \`content\`, não strings puras
-- Trate nomes de tools/resources desconhecidos com erros explícitos
-- Nunca exponha secrets via outputs de tools ou conteúdo de resources
-- Documente cada argumento no campo \`description\` do inputSchema
-
-## Contexto (${graphrag})
-${mcp ? `- Consulte o DARE MCP Server em http://localhost:3000 para contexto do projeto` : `- Use DARE/BLUEPRINT.md como fonte de verdade`}
-
-## Ralph Loop (obrigatório antes de DONE)
-1. Build (npm run build / python -m py_compile)
-2. Test (npm test / pytest)
-3. Inspect — teste as tools com MCP Inspector
-4. Só marque DONE se os 3 passos passarem
+    return `# DARE Framework — MCP Server
+
+## Metodologia
+Você é o Claude Code implementando um servidor MCP com o método DARE:
+- **D**esign: Quais tools/resources/prompts este servidor expõe?
+- **A**rchitect: Schemas das tools, estratégia de transport e auth
+- **R**eview: Teste cada tool com MCP Inspector antes de marcar DONE
+- **E**xecute: Implementação task a task com Ralph Loop
+
+## Configuração do MCP Server
+- Transport: ${mcpTransport}
+- Features: ${featuresStr}
+${transportNote}
+
+## Regras Fundamentais
+- Leia \`DARE/BLUEPRINT.md\` antes de implementar qualquer tool
+- Atualize \`DARE/TASKS.md\` ao concluir cada task
+- Nunca pule o Ralph Loop antes de marcar DONE
+- Teste cada tool com MCP Inspector antes de marcar DONE
+- Valide que o \`inputSchema\` corresponde exatamente ao handler
+
+${langRules}
+
+## Boas Práticas MCP
+- Nomes de tools em snake_case e descritivos
+- Retorne arrays de \`content\`, não strings puras
+- Trate nomes de tools/resources desconhecidos com erros explícitos
+- Nunca exponha secrets via outputs de tools ou conteúdo de resources
+- Documente cada argumento no campo \`description\` do inputSchema
+
+## Contexto (${graphrag})
+${mcp ? `- Consulte o DARE MCP Server em http://localhost:3000 para contexto do projeto` : `- Use DARE/BLUEPRINT.md como fonte de verdade`}
+
+## Ralph Loop (obrigatório antes de DONE)
+1. Build (npm run build / python -m py_compile)
+2. Test (npm test / pytest)
+3. Inspect — teste as tools com MCP Inspector
+4. Só marque DONE se os 3 passos passarem
 `;
 }
 export function generateClaudeCommands(structure) {
     const isMcp = structure === 'mcp-server';
     return {
-        'dare-design.md': `# /dare-design
-
-Gera ou atualiza o \`DARE/DESIGN.md\` a partir de uma descrição.
-
-## Como usar
-\`/dare-design Quero uma API REST de autenticação com JWT e refresh token\`
-
-## O que fazer
-1. Leia o contexto atual do projeto (package.json, estrutura de pastas)
-2. Se \`DARE/DESIGN.md\` já existir, leia-o antes de atualizar
-3. Crie ou atualize \`DARE/DESIGN.md\` com:
-   - **Descrição** do que será construído
-   - **Objetivos** (checkboxes mensuráveis)
-   - **Restrições** técnicas e de negócio
-   - **Critérios de sucesso** verificáveis
-4. Confirme com o usuário antes de prosseguir para blueprint
-
-$ARGUMENTS
+        'dare-design.md': `# /dare-design
+
+Gera ou atualiza o \`DARE/DESIGN.md\` a partir de uma descrição.
+
+## Como usar
+\`/dare-design Quero uma API REST de autenticação com JWT e refresh token\`
+
+## O que fazer
+1. Leia o contexto atual do projeto (package.json, estrutura de pastas)
+2. Se \`DARE/DESIGN.md\` já existir, leia-o antes de atualizar
+3. Crie ou atualize \`DARE/DESIGN.md\` com:
+   - **Descrição** do que será construído
+   - **Objetivos** (checkboxes mensuráveis)
+   - **Restrições** técnicas e de negócio
+   - **Critérios de sucesso** verificáveis
+4. Confirme com o usuário antes de prosseguir para blueprint
+
+$ARGUMENTS
 `,
-        'dare-blueprint.md': `# /dare-blueprint
-
-Gera o \`DARE/BLUEPRINT.md\`, \`DARE/dare-dag.yaml\` e \`DARE/TASKS.md\` a partir do DESIGN.md.
-
-## Como usar
-\`/dare-blueprint\`
-
-## O que fazer
-1. Leia \`DARE/DESIGN.md\` — obrigatório
-2. Gere \`DARE/BLUEPRINT.md\` com:
-   - Stack tecnológico detalhado
-   - Módulos e responsabilidades
-   - Contratos de API (endpoints, schemas)
-   - Modelo de dados
-   - Decisões arquiteturais justificadas
-3. Gere \`DARE/dare-dag.yaml\` com tasks em grafo de dependências
-4. Gere \`DARE/TASKS.md\` com tabela de status
-5. **Aguarde aprovação humana antes de executar qualquer task**
-
-$ARGUMENTS
+        'dare-blueprint.md': `# /dare-blueprint
+
+Gera o \`DARE/BLUEPRINT.md\`, \`DARE/dare-dag.yaml\` e \`DARE/TASKS.md\` a partir do DESIGN.md.
+
+## Como usar
+\`/dare-blueprint\`
+
+## O que fazer
+1. Leia \`DARE/DESIGN.md\` — obrigatório
+2. Gere \`DARE/BLUEPRINT.md\` com:
+   - Stack tecnológico detalhado
+   - Módulos e responsabilidades
+   - Contratos de API (endpoints, schemas)
+   - Modelo de dados
+   - Decisões arquiteturais justificadas
+3. Gere \`DARE/dare-dag.yaml\` com tasks em grafo de dependências
+4. Gere \`DARE/TASKS.md\` com tabela de status
+5. **Aguarde aprovação humana antes de executar qualquer task**
+
+$ARGUMENTS
 `,
-        'dare-execute.md': `# /dare-execute
-
-Executa uma task específica do \`DARE/dare-dag.yaml\` seguindo o Ralph Loop.
-
-## Como usar
-\`/dare-execute task-001\`
-\`/dare-execute task-003 --force\`
-
-## O que fazer
-1. Leia \`DARE/BLUEPRINT.md\` — obrigatório antes de qualquer implementação
-2. Leia a task especificada em \`DARE/dare-dag.yaml\`
-3. Verifique se todas as dependências da task estão com status DONE
-4. Implemente a task
-5. Execute o Ralph Loop:
-   - Build: compile sem erros
-   - Test: todos os testes passando
-   - Lint: sem warnings${isMcp ? '\n   - Inspect: teste com MCP Inspector' : ''}
-6. Atualize o status da task em \`DARE/TASKS.md\` para DONE
-7. Informe o usuário e sugira a próxima task disponível
-
-$ARGUMENTS
+        'dare-execute.md': `# /dare-execute
+
+Executa uma task específica do \`DARE/dare-dag.yaml\` seguindo o Ralph Loop.
+
+## Como usar
+\`/dare-execute task-001\`
+\`/dare-execute task-003 --force\`
+
+## O que fazer
+1. Leia \`DARE/BLUEPRINT.md\` — obrigatório antes de qualquer implementação
+2. Leia a task especificada em \`DARE/dare-dag.yaml\`
+3. Verifique se todas as dependências da task estão com status DONE
+4. Implemente a task
+5. Execute o Ralph Loop:
+   - Build: compile sem erros
+   - Test: todos os testes passando
+   - Lint: sem warnings${isMcp ? '\n   - Inspect: teste com MCP Inspector' : ''}
+6. Atualize o status da task em \`DARE/TASKS.md\` para DONE
+7. Informe o usuário e sugira a próxima task disponível
+
+$ARGUMENTS
 `,
-        'dare-tasks.md': `# /dare-tasks
-
-Exibe o status atual de todas as tasks do projeto.
-
-## Como usar
-\`/dare-tasks\`
-\`/dare-tasks --pending\`
-
-## O que fazer
-1. Leia \`DARE/TASKS.md\` e \`DARE/dare-dag.yaml\`
-2. Exiba uma tabela com: ID, título, status, dependências
-3. Destaque as tasks que estão prontas para execução (dependências satisfeitas)
-4. Calcule e exiba o progresso geral (% concluído)
-
-$ARGUMENTS
+        'dare-tasks.md': `# /dare-tasks
+
+Exibe o status atual de todas as tasks do projeto.
+
+## Como usar
+\`/dare-tasks\`
+\`/dare-tasks --pending\`
+
+## O que fazer
+1. Leia \`DARE/TASKS.md\` e \`DARE/dare-dag.yaml\`
+2. Exiba uma tabela com: ID, título, status, dependências
+3. Destaque as tasks que estão prontas para execução (dependências satisfeitas)
+4. Calcule e exiba o progresso geral (% concluído)
+
+$ARGUMENTS
 `,
     };
 }
 export function generateSharedConfig(projectName) {
-    return `# ${projectName} — DARE Framework
-
-## Design → Architect → Review → Execute
-
-### Estrutura
-- \`DARE/\` — documentação de design e execução
-- \`DARE/EXECUTION/\` — tasks executadas e telemetria
-- \`templates/\` — templates de documentação DARE
-
-### Ralph Loop
-Antes de marcar qualquer task como DONE:
-1. Build ✅
-2. Test ✅
-3. Lint ✅
-
-### Fluxo de Trabalho
-1. \`/generate-design\` — cria DESIGN.md
-2. \`/generate-blueprint\` — cria BLUEPRINT.md
-3. \`/generate-tasks\` — cria TASKS.md
-4. \`/execute-task\` — executa cada task com Ralph Loop
+    return `# ${projectName} — DARE Framework
+
+## Design → Architect → Review → Execute
+
+### Estrutura
+- \`DARE/\` — documentação de design e execução
+- \`DARE/EXECUTION/\` — tasks executadas e telemetria
+- \`templates/\` — templates de documentação DARE
+
+### Ralph Loop
+Antes de marcar qualquer task como DONE:
+1. Build ✅
+2. Test ✅
+3. Lint ✅
+
+### Fluxo de Trabalho
+1. \`/generate-design\` — cria DESIGN.md
+2. \`/generate-blueprint\` — cria BLUEPRINT.md
+3. \`/generate-tasks\` — cria TASKS.md
+4. \`/execute-task\` — executa cada task com Ralph Loop
 `;
 }
 export function generateClaudeSettings(stack) {