convez.ai
SkillProdutividadeIniciante
solntseff

Consilium — Conselho de Decisão com 5 Advisors

solntseff/consilium

5 consultores independentes analisam sua decisão e entregam veredicto honesto

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill que executa qualquer decisão através de 5 consultores independentes com estilos de pensamento distintos, que se desafiam anonimamente e entregam um veredicto sintetizado e honesto. Elimina o viés de concordância do Claude ao forçar análise multifacetada e confrontação de perspectivas.

Instalação

$npx convez-ai solntseff/consilium

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
1
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
v1.0.0
Issues abertas
0
Contribuidores
1

Casos de uso

  • Validar decisões estratégicas de negócio antes de executar
  • Avaliar pivôs ou mudanças significativas de rumo
  • Questionar premissas e encontrar blind spots em planos

Veredito

Consilium resolve um problema real: o viés de concordância do Claude em decisões críticas. A abordagem de múltiplos advisors com cross-review anônimo é inteligente e operacional, entregando recomendações acionáveis sem conforto vazio. Excelente para decisões que requerem perspectivas conflitantes.

README

A tradução para o português deste README ainda está na fila.

🧠 Consilium — Claude Council Skill

Stop trusting Claude's first answer. Run any decision through 5 advisors who think differently, challenge each other anonymously, and deliver a verdict you can actually act on.

License: MIT Works in Claude.ai Works in Claude Code Language: Any


The Problem

Claude is incredibly agreeable. Ask "should I launch this product?" and it finds 5 reasons why you should. Ask "is this product a bad idea?" and it finds 5 reasons why it is. Same product, different framing, opposite answers.

That's fine for writing emails. It's dangerous for making decisions.

A Stanford study found that AI assistants affirm user decisions 49% more often than human respondents — even when the decision is clearly wrong. Every time you ask Claude for advice on a hard decision, there is roughly a coin-flip chance it is quietly nodding along to make you feel good.


The Solution

Consilium runs your question through 5 advisors with fundamentally different thinking styles. They analyze independently, cross-review each other anonymously, and a chairman synthesizes a final verdict with a concrete recommendation.

No cheerleading. No comfort. Just honest analysis.


The 5 Advisors

AdvisorRole
🔴The CriticFinds everything that can go wrong
🔵The Principles AdvisorIgnores details — asks what you're actually trying to achieve
🟢The OptimistFinds the opportunity you're missing
⚪The Detached ObserverCold, dry, takes nothing at face value
🟡The ExecutorLooks only at the next concrete step

How It Works

Your question
      │
      ▼
┌─────────────────────────────────────┐
│  PHASE 1 — Independent Analyses     │
│  5 advisors work in isolation       │
│  No peeking at each other's work    │
└─────────────────────────────────────┘
      │
      ▼
┌─────────────────────────────────────┐
│  PHASE 2 — Anonymous Cross-Review   │
│  Each advisor critiques the others  │
│  Names hidden to prevent groupthink │
└─────────────────────────────────────┘
      │
      ▼
┌─────────────────────────────────────┐
│  PHASE 3 — Chairman's Verdict       │
│  ✅ Consensus                        │
│  ⚡ Key disagreement                 │
│  📌 Recommendation                   │
│  ⚠️  What you lose                   │
│  🚀 First step                       │
└─────────────────────────────────────┘

Trigger Word

consilium

Add it to any question:

"Should I hire a VA or build an automation first? consilium"

Or lead with it:

"consilium: I'm thinking of pivoting my business from X to Y"


Installation

⚡ Option 1 — One-line install via Claude (easiest)

Open a new chat in Claude and paste this:

Please install this Claude skill for me.
The SKILL.md file is at: https://raw.githubusercontent.com/solntseff/consilium/main/SKILL.md
Set it up so I can start using "consilium" as a trigger.

🖥️ Option 2 — Claude.ai Projects (recommended for regular use)

  1. Open claude.ai
  2. Go to Projects → create a new project (e.g. "Strategic Decisions")
  3. Click Project Instructions
  4. Copy the full prompt from SKILL.md and paste it in
  5. Done — consilium now works automatically in every chat inside that project

💬 Option 3 — Any single chat

  1. Open SKILL.md
  2. Copy the entire file content
  3. Paste it at the very start of any Claude conversation
  4. Works for that session only

🖥️ Option 4 — Claude Code (terminal)

curl -o ~/.claude/skills/consilium.md \
  https://raw.githubusercontent.com/solntseff/consilium/main/SKILL.md

Then trigger with consilium in any Claude Code session.


🗂️ Option 5 — Claude Cowork (desktop app)

  1. Download SKILL.md
  2. Open Claude Cowork → Skills
  3. Upload the file
  4. Use consilium as the trigger phrase

When to Use It

✅ Good questions for Consilium:

  • Should I launch now or wait another month?
  • Which of these two options is stronger?
  • My client wants a 30% discount for a long-term contract — should I take it?
  • Am I solving the right problem here?
  • Should I pivot from X to Y?
  • Should I hire someone or automate this first?

❌ Skip Consilium for:

  • Factual questions ("what's the capital of France?")
  • Creative tasks ("write me a post")
  • Simple processing ("summarize this article")

The rule: use Consilium when there's genuine uncertainty and the cost of a bad call is high. If you already know the answer and just want validation — the council will likely tell you things you don't want to hear. That's the point.


Language

Consilium automatically responds in the language you write in — English, Hebrew, Russian, or any other. No setup needed.


Example Output Structure

🔔 Consilium convened. Question received. Summoning advisors...

🔴 THE CRITIC
[Risks and failure points]

🔵 THE PRINCIPLES ADVISOR
[Strategic reframe of the real goal]

🟢 THE OPTIMIST
[Overlooked opportunities]

⚪ THE DETACHED OBSERVER
[Cold objective assessment]

🟡 THE EXECUTOR
[One concrete next step]

── CROSS-REVIEW ──
[Each advisor anonymously comments on the others]

── CHAIRMAN'S VERDICT ──
✅ Consensus: ...
⚡ Key disagreement: ...
📌 Recommendation: ...
⚠️  What you lose: ...
🚀 First step: ...

FAQ

Does this work in the free Claude plan? Yes. Consilium is a prompt — it works on any plan that supports Projects or custom instructions.

Does it work in languages other than English? Yes. Write your question in any language and Consilium responds in the same language.

Can I customize the advisor roles? Absolutely. Open SKILL.md and edit the advisor descriptions to fit your context — industry, team, use case.

Is this the same as LLM Council? It's inspired by the same methodology but uses a different set of advisor roles designed specifically for business decision-making, not code review.


Credits

Designed and built by Alexander Solntseff — WEBSMODE, specialists in business automation and AI integration.

Based on the LLM Council methodology by Andrej Karpathy, adapted by Ole Lehmann.


License

MIT — free to use, share, and modify.

Ver o original no GitHub

Bom para

Relacionados