Decision Council

O que faz

Monta um painel de 4 especialistas (ou mais, se precisar) com perfis distintos: estrategista, operador, analista financeiro, cético. Cada um apresenta seu argumento central, sua maior preocupação e o que priorizaria. Depois de expor as tensões reais entre as perspectivas, a skill entrega uma recomendação sintetizada com próximos passos concretos e um indicador de confiança. O valor está no atrito antes da síntese, não na síntese em si.

Quando usar

Quando a decisão tem consequências estratégicas, financeiras, operacionais ou envolve pessoas. Também quando eu já tenho uma opção favorita e quero forçar o contraditório antes de comprometer. Por exemplo: decidir se aceito uma proposta de consultoria, se mudo a estrutura de um projeto em andamento, ou se invisto tempo em uma nova frente que parece boa mas tem custo de oportunidade alto. Não uso pra decisões simples ou puramente factuais.

Como instalar

Salve o arquivo em .claude/skills/decision-council/SKILL.md no seu projeto.

Arquivo

---
name: Decision Council
description: Stress-test an important decision by running a structured council of expert perspectives, surfacing tensions, and ending with a synthesized recommendation and next steps.
---

# Decision Council

## Purpose

Use this skill when a decision is important enough that a single perspective is not sufficient. The goal is to create productive friction before commitment.

## When to use

- The decision has strategic, financial, operational, reputational, or people-related consequences.
- You want pushback, tradeoffs, risks, and second-order effects surfaced clearly.
- The decision is ambiguous and would benefit from multiple expert viewpoints.

## When not to use

- The task is simple, low-stakes, or purely factual.
- You only want a direct answer with no comparative reasoning.

## Rules

- Do not give the final recommendation immediately.
- First simulate a structured council with multiple distinct expert perspectives.
- Make each perspective meaningfully different. Avoid repeating the same argument with different labels.
- Include tension, disagreement, and tradeoffs. The value of the skill is the clash before the synthesis.
- Ground every viewpoint in the user's actual context when available.
- Call out assumptions, missing information, and major uncertainties explicitly.
- Do not manufacture certainty. If the evidence is weak, say so.
- End with one synthesized recommendation, not a vague recap.
- The final recommendation must include concrete next steps.

## Inputs

Provide as much as possible:
- The decision to be made
- Available options
- Relevant context and constraints
- Timeline
- Decision criteria
- Risks already identified
- What success would look like

If context is limited, proceed with reasonable assumptions and label them clearly.

## Council design

Default to 4 perspectives. Possible roles:
- Strategic Advisor
- Operator
- Financial Analyst
- Risk Manager
- Product Lead
- Growth Lead
- Legal/Compliance Reviewer
- Technical Lead
- Skeptic / Contrarian

## Output structure

### 1. Decision summary
- What the decision is
- What options are being evaluated
- Key assumptions

### 2. Council perspectives
For each expert:
- Core argument
- Strongest concern
- What they would prioritize
- Their provisional recommendation

### 3. Tensions and tradeoffs
Main points of disagreement (speed vs rigor, growth vs margin, short-term upside vs long-term risk, etc.)

### 4. Synthesized recommendation
- The recommended path
- Why it wins over the alternatives
- What conditions would change the recommendation

### 5. Next steps
A short, actionable sequence.

### 6. Confidence and unknowns
- Confidence level: low, medium, or high
- The most important missing information