genesis-planning-nirvana v3.0.0

🌌 Compreensão Celestial de Requisitos

Squad NIRVANA v3 que transforma queries vagas em especificações primordiais cristalinas — alimenta agentes orquestradores com contexto perfeito para planejamento de projetos.

🎯 O Problema Que Resolve

Antes:

User: "Quero criar um SaaS de gestão de projetos"

Orquestrador: 🤔 Que tipo? Para quem? Que escala? Que budget?
              → Começa sem contexto suficiente
              → Decisões técnicas baseadas em guesswork
              → 60% de chance de escolher stack errada

Depois (com genesis-planning-nirvana):

User: "Quero criar um SaaS de gestão de projetos"

genesis-planning-nirvana:
  → Escava intenção real (não apenas pedido superficial)
  → Pesquisa 2025-2026 tech landscape (web_search)
  → Descobre 5-7 alternativas viáveis por decisão
  → Analisa constraints, risks, patterns, resources
  → Sintetiza em primordial_specification.json

Orquestrador: 📋 Recebe spec cristalina:
              - Intent: "Project mgmt for dev teams, Git integration, real-time collab"
              - Stack recomendada: Next.js + PostgreSQL (research de 7 alternativas)
              - Constraints: budget <$500/mês, timeline 8 semanas MVP
              - Risks: Git provider integration complexity (mitigação: usar Octokit)
              - Resources: 1 dev full-stack, 6-8 semanas, $19/mês infra
              → ready_to_plan: true ✅

🏗️ Arquitetura

Pipeline de 8 Agentes (Sequential com Gates)

genesis-intent-archaeologist       (Escava intenção REAL)
  ↓ [gate: intent clarity >= 80%]
  
genesis-domain-cartographer       (Mapeia domínio completo)
  ↓ [gate: domain mapped]
  
genesis-tech-oracle              (Research 2025-2026 tech)
  ↓ [gate: min 5 alternatives]
  
genesis-constraint-analyzer      (Mapeia constraints & trade-offs)
  ↓ [gate: constraints identified]
  
genesis-risk-prophet             (Prediz risks & mitigações)
  ↓ [gate: risks cataloged]
  
genesis-pattern-weaver          (Identifica patterns aplicáveis)
  ↓ [gate: patterns identified]
  
genesis-resource-estimator      (Estima time/cost/team)
  ↓ [gate: estimates provided]
  
genesis-synthesis-master        (Sintetiza em spec primordial)
  ↓ [self-verify: completeness]
  
  OUTPUT: primordial_specification.json

Harness v3

• ✅ Doom loop protection (max 3 identical → change-strategy)
• ✅ Ralph loop (max 5 retries, fresh context)
• ✅ Filesystem collaboration (

  .artifacts/genesis-planning/

  )
• ✅ Context compaction (key-fields, 4K tokens/agent)
• ✅ Model routing:
  • Orchestrator/reasoning: claude-sonnet-4
  • Workers/fast: gemini-flash-2.0

🧠 Research Foundation

Baseado em 75KB+ de pesquisa sobre estado da arte 2025-2026:

1. Meta-Prompting & Context Engineering

• ACE (Agentic Context Engineering): Contextos como playbooks evolutivos
• Context drift = 65% das falhas (não limites de tokens)
• Plan-and-Execute > ReAct para tarefas estruturadas
• Inference-time scaling: Progresso vem de raciocínio durante execução

2. Claude Code DNA (Leaked Source)

• Tool-first architecture: Modularidade extrema
• Progress streaming: onProgress() atualiza real-time
• Permission system: First-class citizen
• Lazy discovery: ToolSearchTool busca dinâmica
• Parallel prefetch: Heavy ops em paralelo no startup

3. Expert Consultant Pattern

1. Discovery: Entender problema real
2. Research: Mínimo 5 alternativas
3. Analysis: Decision matrix objetiva
4. Recommendation: Campeão + finalistas + trade-offs
5. Execution: Implementar + documentar

📋 Output: primordial_specification.json

Schema Master

{
  "project_dna": {
    "intent": { /* Intent Map */ },
    "domain": { /* Domain Topology */ },
    "tech_landscape": { /* Tech Research + Decision Matrices */ },
    "constraints": { /* Constraint Landscape */ },
    "risks": { /* Risk Map */ },
    "patterns": { /* Pattern Catalog */ },
    "resources": { /* Resource Estimates */ }
  },
  
  "primordial_specification": {
    "what_we_are_building": "Crystal-clear 1-paragraph description",
    "why_it_matters": "Value proposition",
    "who_its_for": "Target users",
    "core_value_proposition": "Unique value",
    "success_looks_like": "Success definition"
  },
  
  "recommended_stack": {
    "frontend": {
      "framework": "Next.js 15",
      "rationale": "Research de 7 alternativas, scores objetivos...",
      "alternatives_considered": ["Remix", "SvelteKit", "Astro"],
      "trade_offs": "Vendor lock-in moderado vs velocidade dev"
    },
    "backend": { /* ... */ },
    "database": { /* ... */ },
    "infrastructure": { /* ... */ }
  },
  
  "execution_roadmap": {
    "phase_1_mvp": {
      "goal": "Validate core value prop with 10 beta users",
      "timeline": "6-8 weeks",
      "milestones": ["M001: Core features", "M002: Beta launch"],
      "validation_criteria": "5/10 users complete onboarding"
    },
    "phase_2_production": { /* ... */ },
    "phase_3_scale": { /* ... */ }
  },
  
  "orchestrator_context": {
    "ready_to_plan": true,
    "confidence_level": "high",
    "blocker_questions": [],
    "green_lights": [
      "Stack validated with research",
      "MVP scope clear",
      "Budget constraints met"
    ],
    "yellow_flags": [
      "Real-time collab complexity (mitigated)"
    ],
    "red_flags": []
  }
}

🚀 Usage

Via Squad System

# Activate squad
pi squad activate genesis-planning-nirvana

# Run comprehension flow
pi squad workflow \
  --squad genesis-planning-nirvana \
  --workflow genesis-comprehension-flow \
  --context "Quero criar um SaaS de gestão de projetos para times de desenvolvimento"

# Output: .artifacts/genesis-planning/primordial_specification.json

Integration with GSD-PI

# Feed to GSD auto-mode
pi /gsd auto --context-from .artifacts/genesis-planning/primordial_specification.json

✅ Success Criteria

Squad é bem-sucedido se:

• ✅ Transforma queries vagas → specs cristalinas
• ✅ Research é data-driven (mínimo 5 alternativas, sources citadas)
• ✅ Decision matrices são objetivas (scores numéricos, não feelings)
• ✅ Trade-offs são honestos (não marketing, realidade)
• ✅ Output JSON é válido e completo
• ✅ Orquestrador consegue começar planejamento imediatamente
• ✅ 90%+ de sessões terminam com ready_to_plan=true
• ✅ Tempo médio de compreensão < 5 minutos

📚 Agents Reference

Ver

agents/

🔗 Integration Points

Input

• Format: Free text (user query)
• Source: CLI, API, webhook, etc.
• Example: "Quero criar um SaaS..."

Output

• Format: primordial_specification.json
• Location:

  .artifacts/genesis-planning/

• Schema: Ver

  schemas/primordial_specification.schema.json

Consumers

• GSD-PI auto-mode orchestrator
• nirvana-backend squad
• awwwards-singularity-studio squad
• Qualquer squad de implementação

Provides to GSD Ecosystem

• Seeds CONTEXT.md inicial
• Popula REQUIREMENTS.md
• Sugere milestone structure
• Define tech stack com rationale

🎓 Research Citations

Este squad incorpora insights de:

1. META_INSTRUCTIONS_AI_AGENTS_PLANNING_2026.md (30KB)

   • ACE framework (ICLR 2026)
   • Context engineering patterns
   • Plan-and-Execute vs ReAct
   • Inference-time scaling trends

2. CLAUDE_CODE_SECRETS_DNA_LEAKED.md (25KB)

   • Tool-first architecture (512K linhas source)
   • Permission systems
   • Progress streaming patterns
   • Parallel execution strategies

3. expert-consultant-pattern.md (20KB)

   • 5-phase discovery loop
   • Decision matrix frameworks
   • Single vs multi-select patterns
   • Recommendation presentation

📊 Performance Targets

🛠️ Development Status

Version: 3.0.0
Harness: v3 (doom loop, ralph loop, fs-collab, traces, compaction)
Status: ⚠️ In Development

Roadmap:

• Architecture design
• Research foundation (75KB)
• Squad.json structure
• Agent implementations (8/8)
• Workflow definition
• Schema validation
• Integration testing
• Production deployment

📝 License

MIT

Created: 2026-04-01
Foundation: 75KB de pesquisa sobre meta-instructions, context engineering, e Claude Code DNA
Purpose: Elevar compreensão de requisitos a nível NIRVANA celestial ✨