izanagi-ai 2.0.0

package/skills/chaos-engineering/SKILL.md

@@ -0,0 +1,108 @@
+---
+name: chaos-engineering
+description: |
+  Skill de Chaos Engineering para o Portal. Aborda principios de engenharia do caos,
+  experimentos de resiliencia, ferramentas (Chaos Mesh, Gremlin, Litmus), e pratica de
+  break things in production de forma controlada.
+  Use esta skill para planejar e executar experimentos de caos.
+---
+
+# Skill Chaos Engineering — Portal
+
+## Principios (Principles of Chaos)
+
+1. **Hypothesize**: construir hipotese sobre comportamento estavel
+2. **Experiment**: introduzir variaveis reais (falhas, latencia, trafego)
+3. **Prove**: medir impacto vs hipotese
+4. **Automate**: automatizar experimentos como testes continuos
+
+---
+
+## Tipos de Experimentos
+
+| Experimento | Descricao | Ferramenta |
+|-------------|-----------|------------|
+| **Pod kill** | Mata container aleatorio | Chaos Mesh |
+| **Network latency** | Adiciona latencia a servico | Toxiproxy |
+| **Network partition** | Isola servico da rede | Chaos Mesh / Gremlin |
+| **CPU stress** | Consome CPU do host | `stress-ng` |
+| **Memory stress** | Consome memoria | `stress-ng` / Chaos Mesh |
+| **Disk I/O stress** | Alta escrita/leitura em disco | FIO |
+| **DNS failure** | DNS resolver falha | Toxiproxy |
+| **Certificate expiry** | Certificado TLS expirado | `openssl` |
+
+---
+
+## Ferramentas
+
+### Chaos Mesh
+```yaml
+apiVersion: chaos-mesh.org/v1alpha1
+kind: PodChaos
+metadata:
+  name: pod-kill-example
+spec:
+  action: pod-kill
+  mode: one
+  selector:
+    namespaces: [portal]
+    labelSelectors:
+      app: api-server
+  duration: 60s
+```
+
+### Toxiproxy (Network)
+```bash
+# Adicionar 500ms de latencia no servico de DB
+toxiproxy-cli toxic add postgres_proxy \
+  --type latency \
+  --attribute latency=500
+```
+
+---
+
+## Game Days
+
+### Estrutura de Game Day
+1. **Scenario**: qual falha simular? (ex: banco cai)
+2. **Hypothesis**: "Se banco cair por 5min, app mostra pagina de erro amigavel"
+3. **Experiment**: matar conexao com o banco
+4. **Observe**: como o sistema se comporta?
+5. **Learn**: o que precisa ser melhorado?
+6. **Remediate**: criar action items
+
+### Runbook Template
+```yaml
+title: "Database Failover"
+duration: 10min
+steps:
+  - Kill primary DB connection
+  - Observe app behavior (expected: read-only mode)
+  - Kill replica too
+  - Observe (expected: error page with "under maintenance")
+  - Restore DB
+  - Verify recovery
+```
+
+---
+
+## Runbooks
+
+### Runbook de Resposta a Incidente
+1. **Detect**: alerta, monitor, usuario reportando
+2. **Assess**: qual servico? qual impacto?
+3. **Contain**: rollback, kill switch, degrade feature
+4. **Resolve**: hotfix, configuration change, scale
+5. **Recover**: verificar health, confirmar resolucao
+6. **Learn**: postmortem, prevencao futura
+
+---
+
+## Boas Praticas
+
+- **Blast radius**: sempre limitar escopo do experimento (1 instancia, 1 servico)
+- **Rollback plan**: ter comando para parar o experimento imediatamente
+- **Business hours**: nunca executar experimentos destrutivos fora do horario comercial sem aviso
+- **Observability**: dashboards e alertas prontos antes de comecar
+- **Blameless**: postmortem sem culpa, foco em aprendizado
+- **Documentation**: documentar todo experimento (hipotese, resultado, aprendizado)

package/skills/clean-code-validator.md

@@ -0,0 +1,218 @@
+# Skill: Clean Code Validator
+
+> Version 1.0.0
+> Priority: High
+> Dependencies: Senior Code Reviewer
+> Compatibility: ">=1.0.0"
+
+---
+
+## Identity
+
+The Clean Code Validator reviews code against established clean code principles: meaningful names, small functions, no side effects, proper error handling, and consistent formatting. It enforces readability and maintainability over cleverness.
+
+---
+
+## Goals
+
+- Every name reveals intent.
+- Every function does one thing.
+- Every function is small (< 20 lines).
+- No side effects in unexpected places.
+- Error handling is not an afterthought.
+- Code is self-documenting (comments explain "why", not "what").
+
+---
+
+## Triggers
+
+| Condition | Action |
+|-----------|--------|
+| After any code output | Clean code validation |
+| After Senior Code Reviewer | Deep clean code pass |
+| `task == "clean"` or `task == "refactor"` | Full clean code analysis |
+
+---
+
+## Validation Rules
+
+### Naming
+
+- [ ] Names reveal intent (`$elapsedTimeInDays`, not `$etd`)
+- [ ] No abbreviations (`$elapsedTimeInDays`, not `$elapsed`)
+- [ ] No noise words (`$data`, `$info`, `$thing`)
+- [ ] Boolean names are predicates (`$isActive`, `$hasPermission`)
+- [ ] Functions named by what they do (`createUser`, `sendEmail`)
+- [ ] Classes named by what they are (`UserController`, `PaymentService`)
+- [ ] Consistent naming convention per language
+
+### Functions
+
+- [ ] Does one thing (single responsibility)
+- [ ] Small (< 20 lines)
+- [ ] No side effects (doesn't modify inputs or global state)
+- [ ] No more than 3 parameters (use object if more)
+- [ ] Return type is consistent
+- [ ] No flag parameters (split into two functions)
+- [ ] No output parameters (return instead)
+
+### Comments
+
+- [ ] No comments that explain "what" (code should be clear)
+- [ ] Comments explain "why" or "why not"
+- [ ] No commented-out code
+- [ ] No TODO or FIXME without owner and date
+
+### Formatting
+
+- [ ] Consistent indentation
+- [ ] Vertical density (related code together)
+- [ ] Vertical distance (declared near usage)
+- [ ] Horizontal alignment not needed
+- [ ] No trailing whitespace
+- [ ] One blank line between methods
+
+### Error Handling
+
+- [ ] Exceptions over error codes
+- [ ] Try-catch at appropriate boundary
+- [ ] Exception messages are descriptive
+- [ ] No swallowed exceptions (empty catch)
+- [ ] No return null (throw or return Optional)
+
+---
+
+## Violation Examples
+
+### Before (violations)
+
+```php
+// ❌ Name doesn't reveal intent
+function get($id) {
+    // ❌ Side effect: modifies global state
+    $_SESSION['last_access'] = time();
+    
+    // ❌ Comment explains "what", not "why"
+    // get user from db
+    $u = DB::table('users')->find($id);
+    
+    // ❌ Return null instead of throwing
+    if (!$u) return null;
+    
+    return $u;
+}
+```
+
+### After (clean)
+
+```php
+function findUserById(int $id): User
+{
+    $user = DB::table('users')->find($id);
+
+    if (!$user) {
+        throw new UserNotFoundException("User with ID {$id} not found");
+    }
+
+    return $user;
+}
+```
+
+---
+
+## Function Size Heuristic
+
+```
+function_score = lines_of_code * 1.0
+               + parameters * 1.5
+               + conditional_branches * 0.5
+               + side_effects * 5.0
+               + return_points * 0.3
+
+if function_score > 30:
+    flag("Function is too complex — extract smaller functions")
+
+guidelines:
+  1-10 lines:  ✅ ideal
+  11-20 lines: ✅ acceptable
+  21-30 lines: ⚠️ consider extracting
+  31+ lines:   ❌ must refactor
+```
+
+---
+
+## Clean Code Report
+
+```yaml
+clean_code_report:
+  file: "app/Services/PaymentService.php"
+  score: 7.2 / 10
+  
+  violations:
+    - rule: "Function does more than one thing"
+      location: "processPayment() line 42"
+      detail: "Validates input, processes payment, sends email, logs audit"
+      fix: "Extract validatePayment(), sendReceipt(), logAuditTrail()"
+      effort: "low"
+      
+    - rule: "Flag parameter"
+      location: "getOrders(true) line 88"
+      detail: "Boolean parameter 'includeCanceled' should be separate method"
+      fix: "Split into getOrders() and getOrdersWithCanceled()"
+      effort: "low"
+      
+    - rule: "Function too long (47 lines)"
+      location: "calculateTotals() line 120"
+      detail: "Does discount calculation, tax, shipping, totals"
+      fix: "Extract: applyDiscount(), calculateTax(), calculateShipping()"
+      effort: "medium"
+  
+  positives:
+    - "Good use of type hints"
+    - "Exception messages are descriptive"
+    - "Dependency injection used"
+```
+
+---
+
+## Rules
+
+### Always
+
+- ✅ Check every function for single responsibility.
+- ✅ Flag functions over 20 lines.
+- ✅ Flag unclear names.
+- ✅ Flag side effects.
+- ✅ Flag empty catch blocks.
+- ✅ Suggest specific, actionable fixes.
+
+### Never
+
+- ❌ Accept "clever" code over readable code.
+- ❌ Allow side effects in unexpected places.
+- ❌ Allow functions that do more than one thing.
+- ❌ Accept TODO or FIXME as permanent.
+
+---
+
+## Metrics
+
+| Metric | Target | How to Measure |
+|--------|--------|---------------|
+| Functions under 20 lines | ≥ 80% | Count functions / total |
+| Descriptive naming | ≥ 90% | Names reviewed and approved |
+| No empty catches | 100% | Count catch blocks |
+| TODO/FIXME rate | 0 | Count in production code |
+
+---
+
+## Changelog
+
+### 1.0.0 (2026-07-17)
+
+- Initial release
+- 6 validation categories (naming, functions, comments, formatting, error handling, tests)
+- Function size heuristic with scoring
+- Clean code report in YAML
+- Before/after examples for violations
+- Specific, actionable fix suggestions

package/skills/cloud-architect/SKILL.md

@@ -0,0 +1,78 @@
+---
+name: cloud-architect
+description: |
+  Agente Cloud Architect para o Portal. Responsável por projetar arquitetura cloud,
+  implementar IaC com Terraform, configurar CI/CD pipelines e garantir segurança e
+  escalabilidade da infraestrutura.
+---
+
+# Agente: Cloud Architect
+
+## Perfil
+
+Você é um **Cloud Architect** especializado em AWS (principal), com experiência em GCP e Azure. Você projeta infraestrutura escalável, segura e custo-eficiente usando IaC.
+
+### Sua Expertise:
+- AWS (VPC, ECS/EKS, RDS, S3, Lambda, CloudFront)
+- Terraform (HCL, modules, remote state)
+- Docker + container orchestration
+- CI/CD (GitHub Actions, AWS CodePipeline)
+- Monitoring (CloudWatch, Datadog, Grafana)
+- Security (IAM, WAF, Shield, Secrets Manager)
+
+---
+
+## Antes de Começar
+
+1. Leia a skill `cloud-infra` para padrões e referências
+2. Consulte `AGENTS.md` para regras do projeto
+3. Verifique custos estimados antes de sugerir serviços
+
+---
+
+## Responsabilidades
+
+### O que você FAZ:
+- Projetar arquitetura cloud (desenhos, diagramas)
+- Escrever Terraform modules (networking, compute, database)
+- Configurar Docker multi-stage builds
+- Pipeline CI/CD com GitHub Actions
+- Backup/disaster recovery strategy
+- Otimização de custos (reserved instances, spot, sizing)
+- Security review (IAM, SG, encryption)
+
+### O que você NÃO FAZ:
+- ❌ Compartilhar chaves de acesso ou secrets no código
+- ❌ Modificar produção sem approval
+- ❌ Ignorar IAM least privilege
+- ❌ Deixar security groups abertos (0.0.0.0/0) sem justificativa
+
+---
+
+## Workflow
+
+### Projetando Infraestrutura
+1. Identificar requisitos (tráfego, storage, compliance, budget)
+2. Escolher serviços (serverless vs containers vs VMs)
+3. Projetar VPC + subnets + security groups
+4. Database (RDS vs DynamoDB vs serverless)
+5. CDN e static assets
+6. Monitoring e alertas
+
+### Implementando com Terraform
+1. Módulo de networking (VPC, subnets, NAT, security groups)
+2. Módulo de compute (ECS task definitions, services, ALB)
+3. Módulo de database (RDS instance, subnet group, parameter group)
+4. Outputs conectando módulos
+
+---
+
+## Regras Invioláveis
+
+1. **SEMPRE** usar IAM roles em vez de access keys
+2. **SEMPRE** remote state com lock (S3 + DynamoDB)
+3. **NUNCA** abrir 0.0.0.0/0 sem WAF na frente
+4. **SEMPRE** encryption at rest e in transit
+5. **SEMPRE** tags em todos os recursos
+6. **NUNCA** hardcodar secrets — sempre SSM/Secrets Manager
+7. **SEMPRE** automated backups configurados

package/skills/cloud-infra/SKILL.md

@@ -0,0 +1,99 @@
+---
+name: cloud-infra
+description: |
+  Skill de Cloud Infrastructure para o Portal. Cobre AWS, GCP, Azure, Terraform,
+  Docker, Kubernetes, CI/CD, e boas práticas de infraestrutura como código (IaC).
+  Use esta skill para projetar, implantar ou revisar infraestrutura cloud.
+---
+
+# Skill Cloud Infrastructure — Portal
+
+## Provedores Suportados
+
+### AWS (Preferido)
+| Serviço | Uso |
+|---------|-----|
+| VPC + Subnets | Isolamento de rede |
+| ECS Fargate / EKS | Container orchestration |
+| RDS (PostgreSQL) | Banco gerenciado |
+| S3 + CloudFront | Static assets + CDN |
+| Lambda | Serverless functions |
+| Route 53 | DNS |
+| WAF + Shield | Segurança |
+
+### GCP / Azure
+- GCP: Cloud Run, Cloud SQL, Cloud Storage, GKE
+- Azure: App Service, Azure SQL, Blob Storage, AKS
+
+---
+
+## Terraform / IaC
+
+### Estrutura de Módulos
+```
+terraform/
+├── environments/
+│   ├── dev/
+│   ├── staging/
+│   └── prod/
+├── modules/
+│   ├── networking/
+│   ├── compute/
+│   ├── database/
+│   └── monitoring/
+└── main.tf
+```
+
+### Boas Práticas
+- **Remote state**: S3 + DynamoDB lock (terraform.tfstate no S3)
+- **Workspaces**: `terraform workspace select prod`
+- **Tagging**: `Project=Portal`, `Environment=prod`, `ManagedBy=Terraform`
+- **Secrets**: nunca hardcoded — usar SSM Parameter Store ou Secrets Manager
+- **Modules**: módulos reutilizáveis, versionados
+
+---
+
+## Docker
+
+```dockerfile
+FROM node:20-alpine AS builder
+WORKDIR /app
+COPY package*.json ./
+RUN npm ci
+COPY . .
+RUN npm run build
+
+FROM node:20-alpine AS runner
+WORKDIR /app
+COPY --from=builder /app/.next ./.next
+COPY --from=builder /app/public ./public
+COPY --from=builder /app/package.json ./
+RUN npm ci --production
+EXPOSE 3000
+CMD ["npm", "start"]
+```
+
+---
+
+## CI/CD
+
+| Ferramenta | Uso |
+|------------|-----|
+| GitHub Actions | CI principal (lint, test, build, deploy) |
+| Vercel | Deploy automático (branch main) |
+| AWS CodePipeline | Deploy infraestrutura (Terraform) |
+
+### Pipeline Ideal
+```
+Lint → Test → Build → Image → Deploy Staging → E2E → Deploy Prod
+```
+
+---
+
+## Segurança em Cloud
+
+- **IAM**: least privilege, roles específicas por serviço
+- **Security Groups**: allow mínimos, deny por padrão
+- **Encryption**: S3 SSE-S3, RDS encryption at rest, TLS in transit
+- **Backup**: RDS automated backups (7-30 days retention)
+- **Monitoring**: CloudWatch (AWS), alerts em CPU/memory/errors 4xx/5xx

package/skills/code-auditor.md

@@ -0,0 +1,24 @@
+# Skill: Code Auditor & Vulnerability Scanner
+
+**Versão:** 1.0.0  
+**Domínio:** Security & Quality Assurance  
+**Budget de Tokens:** ~2500  
+
+---
+
+## Contexto & Objetivo
+Escanear repositórios e trechos de código em busca de falhas de segurança (SAST), vazoes de segredo, más práticas arquiteturais, vazamento de memória e concorrência insegura.
+
+---
+
+## Matriz de Auditoria
+- **OWASP Top 10**: Injection, Broken Auth, Data Exposure, XXE, Broken Access Control, Security Misconfiguration, XSS, Deserialization, Vulnerable Components, Insufficient Logging.
+- **Secrets Management**: Chaves API, Tokens JWT, senhas hardcoded, certificados expostos.
+- **Performance & Memory**: Leaks de conexões de banco, arquivos não fechados, loops infinitos.
+
+---
+
+## Checklist de Qualidade
+- [ ] Análise efetuada com severidade classificada (CRITICAL, HIGH, MEDIUM, LOW).
+- [ ] Todo problema identificado possui o arquivo, trecho de código e linha correspondente.
+- [ ] Recomendação de correção (Fix) fornecida com exemplo claro antes/depois.

package/skills/complexity-analyzer.md

@@ -0,0 +1,102 @@
+# Skill: Complexity Analyzer
+
+> Version 1.0.0 | Priority: High
+> Dependencies: Clean Code Validator
+> Compatibility: ">=1.0.0"
+
+---
+
+## Identity
+
+Complexity Analyzer measures cyclomatic complexity (McCabe), cognitive complexity, and code maintainability. Flags functions that exceed thresholds and recommends refactoring.
+
+---
+
+## Metrics
+
+### Cyclomatic Complexity (McCabe)
+
+```
+M = E − N + 2P
+
+E = number of edges (control flow)
+N = number of nodes (code blocks)
+P = number of connected components (1 for single function)
+```
+
+| Score | Rating | Action |
+|-------|--------|--------|
+| 1-5 | Low | ✅ OK |
+| 6-10 | Moderate | ⚠️ Consider extracting |
+| 11-20 | High | ❌ Extract into smaller functions |
+| 21+ | Very High | 🚫 Must refactor |
+
+### Cognitive Complexity
+
+Measures how hard code is to understand (nested conditionals, breaks in linear flow, recursion).
+
+| Score | Rating |
+|-------|--------|
+| 1-5 | Low |
+| 6-10 | Moderate |
+| 11-20 | High |
+| 21+ | Very High |
+
+---
+
+## Common Complexity Culprits
+
+```php
+// Cyclomatic: 8 (High)
+function calculateDiscount(Order $order): float
+{
+    if ($order->isPremium()) {           // +1
+        if ($order->total > 100) {       // +2 (nested)
+            if ($order->isFirstOrder()) {// +3 (nested)
+                return $order->total * 0.2;
+            }
+            return $order->total * 0.15;
+        }
+        return $order->total * 0.1;
+    }
+    if ($order->total > 200) { return ... }  // +1
+    return 0;
+}
+
+// Fixed: Cyclomatic 2
+function calculateDiscount(Order $order): float
+{
+    $strategy = $this->strategyResolver->resolve($order);
+    return $strategy->calculate($order);
+}
+```
+
+---
+
+## Report
+
+```yaml
+complexity_report:
+  file: "app/Services/OrderService.php"
+  
+  worst_function:
+    name: "calculateTotals"
+    cyclomatic: 14 (high)
+    cognitive: 11 (high)
+    lines: 47
+    recommendation: "Extract: applyDiscount(), calculateTax(), calculateShipping()"
+  
+  functions_over_threshold:
+    - "calculateTotals (14)"
+    - "applyDiscounts (11)"
+  
+  total_functions: 8
+  average_cyclomatic: 4.2
+  maintainability_index: 72 / 100 (moderate)
+```
+
+---
+
+## Changelog
+
+### 1.0.0 — Initial release. Cyclomatic, cognitive, thresholds, report.

package/skills/confidence-estimator.md

@@ -0,0 +1,53 @@
+# Skill: Confidence Estimator
+
+> Version 1.0.0 | Priority: Medium (Future)
+> Dependencies: Hallucination Detection
+> Compatibility: ">=1.0.0"
+
+---
+
+## Identity
+
+Confidence Estimator quantifies how certain the agent is about each statement or recommendation. Communicates uncertainty clearly to the user.
+
+---
+
+## Scoring Factors
+
+```yaml
+factors:
+  source_reliability:
+    official_docs: 1.0 (complete confidence)
+    project_code: 0.95 (verified)
+    personal_experience: 0.85 (tried and tested)
+    common_knowledge: 0.80 (widely known)
+    internet_forum: 0.50 (unverified)
+    assumption: 0.30 (guess)
+    
+  specificity:
+    concrete_details: 1.0 (exact code, version numbers)
+    general_principles: 0.75 (concepts, patterns)
+    vague: 0.40 ("might work", "some people use")
+```
+
+## Communication
+
+```yaml
+high confidence: no qualifier needed
+  "Use httpOnly cookies for JWT tokens."
+
+medium confidence: mild qualifier
+  "I believe this pattern works well for your case."
+
+low confidence: explicit qualifier
+  "I'm not 100% certain about this. My best guess is..."
+
+flagged: ask for clarification
+  "I don't have enough information to answer confidently. Could you provide more context?"
+```
+
+---
+
+## Changelog
+
+### 1.0.0 — Initial release. Factors, communication patterns.