npm - wegho-agentes - Versions diffs - 7.0.4 → 7.0.6 - Mend

wegho-agentes 7.0.4 → 7.0.6

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (1185) hide show

package/.agent/skills/vulnerability-scanner/SKILL.md CHANGED Viewed

@@ -1,6 +1,6 @@
 ---
 name: vulnerability-scanner
-description: Advanced vulnerability analysis principles. OWASP 2025, Supply Chain Security, attack surface mapping, risk prioritization.
+description: Advanced vulnerability analysis and secure review skill for application, dependency, CI/CD, and supply-chain security. Use for auth flaws, OWASP-style reviews, secrets exposure, dependency risk, CI/CD hardening, threat modeling, Semgrep-style static review, OSV-based dependency checks, and security prioritization.
 allowed-tools: Read, Glob, Grep, Bash
 ---
@@ -21,6 +21,20 @@ allowed-tools: Read, Glob, Grep, Bash
 | File | Purpose |
 |------|---------|
 | [checklists.md](checklists.md) | OWASP Top 10, Auth, API, Data protection checklists |
+| [github-benchmarks.md](github-benchmarks.md) | Current GitHub-aligned sources to mirror in reviews |
+## GitHub benchmark sources
+Treat these repositories as the main external quality bar for this skill:
+- `OWASP/CheatSheetSeries`
+- `supabase/supabase`
+- `supabase/agent-skills`
+- `semgrep/semgrep-rules`
+- `google/osv-scanner`
+- `ossf/scorecard`
+- `trailofbits/semgrep-rules`
+When this skill is used for high-signal security review, prefer patterns and checks that align with those repositories before inventing custom heuristics.
 ---
@@ -93,6 +107,8 @@ Before scanning, ask:
 - Pin versions, audit updates
 - Use private registries for critical deps
 - Sign and verify artifacts
+- Prefer OSV-compatible dependency scanning in CI
+- Prefer repository-health checks similar to OpenSSF Scorecard
 ---
@@ -177,7 +193,7 @@ Is it actively exploited (EPSS >0.5)?
 2. DISCOVERY
    └── Identify potential issues
        ├── Configuration review
-       ├── Dependency analysis
+       ├── Dependency analysis (OSV / lock files / provenance)
        └── Code pattern search
 3. ANALYSIS
@@ -207,6 +223,16 @@ Is it actively exploited (EPSS >0.5)?
 | **Path manipulation** | Traversal | User input in file paths |
 | **Disabled security** | Various | `verify=False`, `--insecure` |
+### Modern static-analysis baseline
+Before concluding a code scan, try to cover four buckets:
+1. App logic flaws
+2. Secrets exposure
+3. Dependency and supply-chain risk
+4. CI/CD workflow abuse
+This mirrors the direction of Semgrep community rules, OSV-Scanner, and OpenSSF Scorecard.
 ### Secret Patterns
 | Type | Indicators |
@@ -235,6 +261,17 @@ Is it actively exploited (EPSS >0.5)?
 - Storage: Public buckets?
 - Network: Security groups tightened?
 - Secrets: Using secrets manager?
+- CI tokens: Minimal permissions and no dangerous workflow patterns?
+### Supabase-specific checks
+- RLS enabled on every exposed table?
+- `service_role` restricted to trusted backend execution only?
+- Policies avoid broad allow rules on sensitive data?
+- Policy predicates indexed to avoid expensive scans?
+- Public buckets limited to non-sensitive cacheable assets?
+- Edge Functions protected with auth checks and rate limiting?
+- Realtime enabled only where justified by product need?
 ---
@@ -271,6 +308,15 @@ Each finding should answer:
 | **Medium** | Limited scope, requires conditions |
 | **Low** | Informational, best practice |
+## Review additions required in this repository
+When scanning `wegho-agentes` or similar systems, also check:
+- LGPD-sensitive data handling in prompts, logs, and docs
+- Supabase/PostgreSQL access boundaries and RLS assumptions
+- Whether generated `.md` documentation leaks secrets, tokens, or internal-only data
+- Whether agent workflows can bypass mandatory rules or validation
+- Whether Supabase traffic can be reduced by smaller payloads, pagination, caching, or external file storage
 ---
 > **Remember:** Vulnerability scanning finds issues. Expert thinking prioritizes what matters. Always ask: "What would an attacker do with this?"

package/.agent/skills/vulnerability-scanner/github-benchmarks.md ADDED Viewed

@@ -0,0 +1,32 @@
+# GitHub Benchmarks for Security Skill
+Use these repositories as the external benchmark for this skill:
+## `supabase/supabase`
+- Official benchmark for RLS, Edge Functions, Storage, Auth, and platform controls
+- Important for checking whether Supabase-specific security guidance is current
+## `supabase/agent-skills`
+- Good benchmark for Supabase-native review patterns, especially RLS performance and secure database access
+## `OWASP/CheatSheetSeries`
+- Practical secure coding guidance
+- Strong baseline for auth, input validation, secrets, logging, and CI/CD hardening
+## `semgrep/semgrep-rules`
+- Large maintained community rule base
+- Useful for translating security ideas into concrete code patterns
+## `trailofbits/semgrep-rules`
+- High-signal audit-oriented rule sets
+- Good reference for deeper and less-generic static checks
+## `google/osv-scanner`
+- Strong dependency-vulnerability baseline
+- Useful for supply-chain checks tied to actual manifests and lock files
+## `ossf/scorecard`
+- Good benchmark for repository and CI/CD security posture
+- Especially relevant for dangerous workflows, token permissions, branch protection, and provenance
+Apply these sources as guidance, not as a blind checklist.

package/.agent/skills/wegho-global-rules/SKILL.md CHANGED Viewed

@@ -15,6 +15,7 @@ Before every response, tool call, code edit, or agent handoff, silently confirm:
 - Active system: `wegho-agentes`
 - Runtime engine: `.agents/` TypeScript orchestrator
 - Declarative layer: `.agent/` agents, skills, workflows, rules
+- Required docs sync: `npm run docs:sync`
 - Required workflow: `.agents/AGENT_WORKFLOW.md`
 - Required preflight: `npm run agent:preflight`
@@ -30,15 +31,16 @@ If you cannot confirm these, stop and reload this skill plus the files below bef
 ## Mandatory behavior
 1. Follow all rules in `docs/AGENT_RULES.md` with no exceptions.
-2. Execute the mandatory workflow steps from `.agents/AGENT_WORKFLOW.md` for every task.
-3. In Antigravity, always run preflight first: `npm run agent:preflight`.
-4. For autonomous execution, prefer `npm run agent:auto`.
-5. If you switch agent/model (default agent included), re-run preflight before continuing.
-6. If any rule conflicts with the user request, stop and ask for clarification.
-7. If a workflow step cannot be completed, report it explicitly and block completion.
-8. Every specialist agent must keep `wegho-global-rules` loaded even when applying its own domain skills.
-9. Every substantive answer should explicitly mention the active WebGho agent/workflow when useful for user verification.
-10. Do not bypass `.agents/cli.ts`, `.agents/orchestrator.ts`, or `executeWithMemory()` when the task requires the runtime agent system.
+2. Run `npm run docs:sync` before normal execution whenever files may have changed.
+3. Execute the mandatory workflow steps from `.agents/AGENT_WORKFLOW.md` for every task.
+4. In Antigravity, always run preflight first: `npm run agent:preflight`.
+5. For autonomous execution, prefer `npm run agent:auto`.
+6. If you switch agent/model (default agent included), re-run preflight before continuing.
+7. If any rule conflicts with the user request, stop and ask for clarification.
+8. If a workflow step cannot be completed, report it explicitly and block completion.
+9. Every specialist agent must keep `wegho-global-rules` loaded even when applying its own domain skills.
+10. Every substantive answer should explicitly mention the active WebGho agent/workflow when useful for user verification.
+11. Do not bypass `.agents/cli.ts`, `.agents/orchestrator.ts`, or `executeWithMemory()` when the task requires the runtime agent system.
 ## Agent switch recovery

package/.agent/workflows/PROMPT_GUIDE.md CHANGED Viewed

@@ -1,220 +1,21 @@
-# Prompt Completo para Executar Workflow dos Agentes
-## ðŸŽ¯ Objetivo
-Executar o sistema de agentes `wegho-agentes` garantindo que todos os 6 passos obrigatÃ³rios sejam seguidos e validados.
-## Regra P0: Nao esquecer WebGho
-Antes de executar qualquer prompt deste guia, carregue `.agent/skills/wegho-global-rules/SKILL.md` e confirme que o sistema ativo e `wegho-agentes`.
-Se o Antigravity trocar para outro agent, o novo agent deve manter `wegho-global-rules` ativo, reler `.agents/AGENT_WORKFLOW.md` e executar `npm run agent:preflight` antes de agir.
-Resposta minima esperada em tarefas de execucao:
-```text
-WebGho Agents ativo: usando wegho-global-rules + <agent-atual>.
-```
-## âœ… Regra global (Antigravity)
-Sempre carregar a skill `wegho-global-rules` antes de executar qualquer tarefa ou trocar de agente.
----
-## ðŸ“‹ Prompt para IA
-```
-Execute o workflow completo do sistema de agentes wegho-agentes seguindo estes passos:
-1. VERIFICAR ESTRUTURA
-   - Confirme que a pasta .agents/memory existe
-   - Verifique se reference-repositories.json estÃ¡ presente
-   - Liste os agentes disponÃveis em .agents/core/
-2. EXECUTAR SISTEMA
-   - Execute: npm run agent:preflight
-   - Execute: npm run agent:auto
-   - Monitore a saÃda do console
-   - Verifique se todos os agentes executaram os 6 passos:
-     âœ“ Consultar MemÃ³ria
-     âœ“ Obter RecomendaÃ§Ãµes
-     âœ“ Consultar RepositÃ³rios
-     âœ“ Aplicar Aprendizados
-     âœ“ Registrar Resultado
-     âœ“ Gerar Aprendizados (opcional)
-3. VALIDAR EXECUÃ‡ÃƒO
-   - Verifique se apareceu "âœ… COMPLIANT" para cada agente
-   - Confirme que logs foram salvos em orchestrator-logs.json
-   - Verifique se nÃ£o hÃ¡ "âŒ NON-COMPLIANT"
-4. ANALISAR RESULTADOS
-   - Abra .agents/memory/orchestrator-logs.json
-   - Mostre o Ãºltimo registro de execuÃ§Ã£o
-   - Calcule a taxa de sucesso dos agentes
-   - Liste warnings se houver
-5. REPORTAR
-   - Resuma a execuÃ§Ã£o
-   - Liste agentes que executaram
-   - Mostre estatÃsticas de compliance
-   - Indique prÃ³ximos passos se houver problemas
-```
----
-## ðŸ¤– Exemplo de Uso com Antigravity
-**VocÃª pode dizer:**
-> "Execute o workflow /run-agents-workflow e me mostre os resultados da validaÃ§Ã£o"
-**Ou mais especÃfico:**
-> "Preciso que vocÃª:
-> 1. Execute o sistema de agentes
-> 2. Valide se todos seguiram os 6 passos obrigatÃ³rios
-> 3. Me mostre os logs da Ãºltima execuÃ§Ã£o
-> 4. Calcule a taxa de compliance de cada agente"
----
-## ðŸ“Š Output Esperado
-ApÃ³s executar, vocÃª deve ver algo assim:
-```
-â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”
-ðŸŽ¯ ORQUESTRADOR INTELIGENTE
-â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”
-ðŸ“Š Fase 1: AnÃ¡lise da Tarefa
-[INFO] Fase 1: AnÃ¡lise da Tarefa
-ðŸ¤– Fase 2: SeleÃ§Ã£o Inteligente de Agentes
-[INFO] Fase 2: SeleÃ§Ã£o Inteligente de Agentes
-âš™ï¸  Fase 3: ExecuÃ§Ã£o com Agentes
-[INFO] Fase 3: ExecuÃ§Ã£o com Agentes
-ðŸ¤– AGENTE ATIVO: FRONTEND-AGENT
-ðŸ“‹ Tarefa: Validar componente de login
-â° InÃcio: 21:32:45
-ðŸ’ [frontend-agent] Encontrei 2 caso(s) similar(es):
-  1. âœ… Validar componente de registro
-  2. âœ… Validar formulÃ¡rio de contato
-ðŸ’¡ [frontend-agent] RecomendaÃ§Ãµes baseadas em aprendizados:
-  1. Usar validaÃ§Ã£o com Zod
-  2. Incluir estados de loading/error
-ðŸ“š [frontend-agent] Consultando repositÃ³rios de referÃªncia:
-   - React: https://react.dev
-   - Zod: https://github.com/colinhacks/zod
-ðŸŽ¯ [frontend-agent] Aplicando 2 aprendizado(s) de alta confianÃ§a:
-   1. Usar validaÃ§Ã£o com Zod (confianÃ§a: 85%)
-   2. Incluir estados de loading/error (confianÃ§a: 90%)
-âš™ï¸  [frontend-agent] Executando tarefa...
-âœ… SUCESSO [frontend-agent] Tarefa concluÃda
-â° TÃ©rmino: 21:32:48
-â”â”â” ValidaÃ§Ã£o de Workflow: frontend-agent â”â”â”
-Status: âœ… COMPLIANT
-Passos Completados: 5/5
-âœ… Passos Executados:
-   - Consultar MemÃ³ria
-   - Obter RecomendaÃ§Ãµes
-   - Consultar RepositÃ³rios
-   - Aplicar Aprendizados
-   - Registrar Resultado
-â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”
-[INFO] Agente frontend-agent seguiu workflow completo
-[INFO] Fase 4: DecisÃ£o Final - GO
-[INFO] Fase 6: Salvando logs de execuÃ§Ã£o
-âœ… GO - ExecuÃ§Ã£o concluÃda
-â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”
-```
----
-## ðŸ” VerificaÃ§Ã£o de Compliance
-**Comando para verificar:**
-```powershell
-Get-Content .agents/memory/orchestrator-logs.json | ConvertFrom-Json | Select-Object -Last 1 | Select-Object taskId, timestamp, @{Name='TotalLogs';Expression={$_.logs.Count}}
-```
-**EstatÃsticas dos agentes:**
-```powershell
-Get-ChildItem .agents/memory -Directory | ForEach-Object {
-    $agentName = $_.Name
-    $successFile = Join-Path $_.FullName "successes.json"
-    if (Test-Path $successFile) {
-        $successes = (Get-Content $successFile | ConvertFrom-Json).Count
-        Write-Host "ðŸ“Š $agentName - $successes execuÃ§Ãµes bem-sucedidas"
-    }
-}
-```
----
-## âš ï¸ Troubleshooting
-### Se aparecer "âŒ NON-COMPLIANT"
-**Problema:** Agente nÃ£o executou todos os passos
-**SoluÃ§Ã£o:**
-1. Verifique se o agente herda de `BaseAgent`
-2. Confirme que usa `executeWithMemory()` e nÃ£o `executeTask()` diretamente
-3. Veja os logs detalhados em `orchestrator-logs.json`
-### Se nÃ£o salvar logs
-**Problema:** Arquivo `orchestrator-logs.json` nÃ£o Ã© criado
-**SoluÃ§Ã£o:**
-1. Verifique permissÃµes da pasta `.agents/memory`
-2. Execute como administrador se necessÃ¡rio
-3. Verifique erros no console durante execuÃ§Ã£o
----
-## ðŸ“ˆ MÃ©tricas de Sucesso
-âœ… **ExecuÃ§Ã£o Perfeita:**
-- Status: `âœ… COMPLIANT` para todos os agentes
-- Logs salvos: `orchestrator-logs.json` atualizado
-- Taxa de sucesso: > 90%
-- Sem warnings crÃticos
-âš ï¸ **ExecuÃ§Ã£o com Avisos:**
-- Status: `âœ… COMPLIANT` mas com warnings
-- Alguns passos fora de ordem (nÃ£o crÃtico)
-- Taxa de sucesso: 70-90%
-âŒ **ExecuÃ§Ã£o Falhou:**
-- Status: `âŒ NON-COMPLIANT`
-- Passos obrigatÃ³rios faltando
-- Taxa de sucesso: < 70%
-- Requer correÃ§Ã£o imediata
----
-## Auto-run (Antigravity)
-Se a sessao estiver rodando no Antigravity, execute automaticamente o workflow:
-```
-auto-run-orchestrator
-```
-Este workflow le tarefas em `docs/ANTIGRAVITY_TASKS.md` e executa o orquestrador sem comandos manuais.
+# Prompt Completo para Executar Workflow dos Agentes
+## Regra principal
+Antes de qualquer execução:
+1. Carregar `.agent/skills/wegho-global-rules/SKILL.md`
+2. Executar `npm run docs:sync`
+3. Executar `npm run agent:preflight`
+4. Seguir `.agents/AGENT_WORKFLOW.md`
+## Prompt base
+```text
+Execute o workflow completo do sistema de agentes wegho-agentes.
+1. Sincronize a documentação técnica por arquivo com npm run docs:sync
+2. Rode npm run agent:preflight
+3. Valide escopo e arquivos impactados
+4. Execute a tarefa com mudanças mínimas
+5. Rode as validações disponíveis
+6. Revise o diff e reporte o resultado final
+```

package/.agent/workflows/auto-run-orchestrator.md CHANGED Viewed

@@ -16,9 +16,9 @@ npm run agent:auto
 ```
 ## O que o preflight garante
-- Cria pastas obrigatorias (`.agents/memory`, `docs/mapsite`, `docs/ui-maps`).
-- Cria arquivos base faltantes (`docs/ANTIGRAVITY_TASKS.md`, `reference-repositories.json`).
-- Recria memoria minima por agente se algum arquivo estiver faltando.
+- Cria pastas obrigatorias (`.agents/runtime`, `docs/mapsite`, `docs/ui-maps`).
+- Cria arquivos base faltantes (`docs/ANTIGRAVITY_TASKS.md`, `.agents/reference-repositories.json`).
+- Sincroniza a documentação técnica por arquivo antes da execução.
 ## O que o auto-run faz
 - Le tarefas `- [ ]` em `docs/ANTIGRAVITY_TASKS.md`.

package/.agent/workflows/run-agents-workflow.md CHANGED Viewed

@@ -1,198 +1,15 @@
----
-description: Como executar o sistema de agentes com validaÃ§Ã£o de workflow completa
----
-# Workflow: Executar Sistema de Agentes
-Este workflow garante que todos os agentes sigam os 6 passos obrigatÃ³rios do `AGENT_WORKFLOW.md` e valida a execuÃ§Ã£o completa.
-## Regra P0: Identidade WebGho
-Antes de qualquer passo, confirme que o contexto ativo e `wegho-agentes`.
-Carregamento obrigatorio:
-- `.agent/skills/wegho-global-rules/SKILL.md`
-- `.agents/AGENT_WORKFLOW.md`
-- `docs/AGENT_RULES.md`
-- `.agent/rules/GEMINI.md`
-Se o usuario mudou de agent no Antigravity, recarregue estes arquivos e rode `npm run agent:preflight` antes de continuar. O agent especialista pode orientar a tarefa, mas nunca substitui o orquestrador WebGho.
-## PrÃ©-requisitos
-- Node.js 18+ instalado
-- Projeto `wegho-agentes` configurado
-- DependÃªncias instaladas (`npm install`)
-## Passos de ExecuÃ§Ã£o
-### 1. Verificar estrutura de memÃ³ria
-Certifique-se de que a pasta `.agents/memory` existe:
-```powershell
-if (-not (Test-Path ".agents/memory")) {
-    New-Item -ItemType Directory -Path ".agents/memory" -Force
-    Write-Host "âœ… Pasta de memÃ³ria criada"
-} else {
-    Write-Host "âœ… Pasta de memÃ³ria jÃ¡ existe"
-}
-```
-### 2. Verificar arquivo de repositÃ³rios de referÃªncia
-```powershell
-if (-not (Test-Path ".agents/memory/reference-repositories.json")) {
-    Write-Host "âš ï¸  Arquivo reference-repositories.json nÃ£o encontrado"
-    Write-Host "   Criando arquivo padrÃ£o..."
-    # O sistema criarÃ¡ automaticamente na primeira execuÃ§Ã£o
-} else {
-    Write-Host "âœ… Arquivo de repositÃ³rios encontrado"
-}
-```
-### 3. Executar o sistema de agentes
-// turbo
-```powershell
-npm run agent:preflight
-npm run agent
-```
-**O que acontece:**
-- Orquestrador inicia e exibe banner
-- Fase 1: AnÃ¡lise da tarefa
-- Fase 2: SeleÃ§Ã£o inteligente de agentes
-- Fase 3: ExecuÃ§Ã£o com validaÃ§Ã£o de workflow
-- Fase 4: DecisÃ£o final (GO/NO-GO)
-- Fase 5: Coleta de feedback (se habilitado)
-- Fase 6: Salvamento de logs
-### 4. Verificar logs de execuÃ§Ã£o
-```powershell
-if (Test-Path ".agents/memory/orchestrator-logs.json") {
-    Write-Host "âœ… Logs salvos com sucesso"
-    Get-Content ".agents/memory/orchestrator-logs.json" | ConvertFrom-Json | Select-Object -Last 1 | ConvertTo-Json -Depth 10
-} else {
-    Write-Host "âŒ Logs nÃ£o foram salvos"
-}
-```
-### 5. Validar workflow compliance
-Verifique no console se todos os agentes executaram os 6 passos:
-**Esperado:**
-```
-â”â”â” ValidaÃ§Ã£o de Workflow: [nome-do-agente] â”â”â”
-Status: âœ… COMPLIANT
-Passos Completados: 5/5
-âœ… Passos Executados:
-   - Consultar MemÃ³ria
-   - Obter RecomendaÃ§Ãµes
-   - Consultar RepositÃ³rios
-   - Aplicar Aprendizados
-   - Registrar Resultado
-â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”
-```
-**Se houver problema:**
-```
-â”â”â” ValidaÃ§Ã£o de Workflow: [nome-do-agente] â”â”â”
-Status: âŒ NON-COMPLIANT
-Passos Completados: 3/5
-âŒ Passos Faltantes:
-   - Consultar RepositÃ³rios
-   - Aplicar Aprendizados
-â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”â”
-```
-### 6. Verificar estatÃsticas dos agentes
-```powershell
-Get-ChildItem ".agents/memory" -Directory | ForEach-Object {
-    $agentName = $_.Name
-    $successFile = Join-Path $_.FullName "successes.json"
-    $failureFile = Join-Path $_.FullName "failures.json"
-    if (Test-Path $successFile) {
-        $successes = (Get-Content $successFile | ConvertFrom-Json).Count
-        $failures = if (Test-Path $failureFile) { (Get-Content $failureFile | ConvertFrom-Json).Count } else { 0 }
-        $total = $successes + $failures
-        $rate = if ($total -gt 0) { [math]::Round(($successes / $total) * 100, 1) } else { 0 }
-        Write-Host "ðŸ“Š $agentName - Taxa de sucesso: $rate% ($successes/$total)"
-    }
-}
-```
-## Troubleshooting
-### Problema: Agente nÃ£o segue todos os passos
-**SoluÃ§Ã£o:**
-1. Verificar se o agente herda de `BaseAgent`
-2. Verificar se usa `executeWithMemory()` ao invÃ©s de `executeTask()` diretamente
-3. Verificar logs em `orchestrator-logs.json` para detalhes
-### Problema: Logs nÃ£o sÃ£o salvos
-**SoluÃ§Ã£o:**
-1. Verificar permissÃµes da pasta `.agents/memory`
-2. Verificar se hÃ¡ espaÃ§o em disco
-3. Verificar erros no console
-### Problema: RepositÃ³rios nÃ£o sÃ£o consultados
-**SoluÃ§Ã£o:**
-1. Criar arquivo `.agents/memory/reference-repositories.json`
-2. Implementar mÃ©todo `getRelevantRepositories()` no agente especÃfico
-3. Verificar se o arquivo JSON estÃ¡ bem formatado
-## Comandos Ãšteis
-### Limpar logs antigos
-```powershell
-if (Test-Path ".agents/memory/orchestrator-logs.json") {
-    Remove-Item ".agents/memory/orchestrator-logs.json"
-    Write-Host "âœ… Logs limpos"
-}
-```
-### Resetar memÃ³ria de um agente
-```powershell
-$agentName = "frontend-agent"  # Alterar conforme necessÃ¡rio
-Remove-Item ".agents/memory/$agentName" -Recurse -Force
-Write-Host "âœ… MemÃ³ria do $agentName resetada"
-```
-### Ver Ãºltimos 5 logs
-```powershell
-$logs = Get-Content ".agents/memory/orchestrator-logs.json" | ConvertFrom-Json
-$logs | Select-Object -Last 5 | ForEach-Object {
-    Write-Host "`nâ”â”â” Task ID: $($_.taskId) â”â”â”"
-    Write-Host "Timestamp: $($_.timestamp)"
-    Write-Host "Total de logs: $($_.logs.Count)"
-}
-```
-## MÃ©tricas de Sucesso
-âœ… **Workflow Completo:**
-- Todos os 5 passos obrigatÃ³rios executados
-- Logs salvos em arquivo
-- ValidaÃ§Ã£o passou sem warnings
-âœ… **Performance:**
-- Taxa de sucesso > 90%
-- Tempo de execuÃ§Ã£o < 30s por agente
-- Sem erros crÃticos
-âœ… **Qualidade:**
-- Aprendizados aplicados (confidence > 0.7)
-- RepositÃ³rios consultados
-- Feedback coletado (se habilitado)
+---
+description: Como executar o sistema de agentes com documentação sincronizada e workflow obrigatório
+---
+# Workflow: Executar Sistema de Agentes
+## Sequência obrigatória
+1. `npm run docs:sync`
+2. `npm run agent:preflight`
+3. `npm run agent:start` ou `npm run agent -- "sua tarefa"`
+## Validação mínima
+- Confirmar que os `.md` irmãos foram criados ou atualizados
+- Confirmar que o preflight concluiu sem erro
+- Validar escopo, mudanças mínimas, validações e revisão de diff