The Claude Council: Turn One Assistant Into Five Advisors Who Argue First

I make a lot of decisions out loud with Claude — should I take this client, is this architecture overbuilt, am I about to talk myself into something. The problem with asking any single assistant is that it tends to agree with the framing you handed it. A trick I have been using fixes that: instead of one obliging voice, I make Claude answer as a small council of advisors who argue before they agree. Here is the exact setup, and five ways I put it to work.

What the Council actually is

There is no hidden feature here, and that is the appealing part. The Council is just a Claude Project with a set of standing instructions. A Project is a workspace where every chat you open inherits the same instructions and files automatically, so you configure the behaviour once and never paste it again. Name the project, give it the council instructions below, and from then on every conversation inside it answers as five distinct advisors instead of one agreeable assistant.

Setting it up in about two minutes

Open the Projects section in the Claude sidebar and choose New Project. Name it something like The Council. Open the project settings and paste the instructions from the next section into the project instructions field. That is the whole setup. Every new chat you start inside that project now runs as a council; chats outside it behave normally, so you are not changing how the rest of Claude works.

The instructions to paste

The core rule is simple: Claude must never reply as a single voice. For every question it spins up five advisors, each taking a deliberately different angle, then closes with one verdict. Paste a version of this into the project instructions, telling Claude that for every question it should respond as the following five advisors and then reconcile them:

• The Contrarian — goes straight at the weakest link in your reasoning and says what no one else will
• The First-Principles Thinker — ignores how you phrased the question and solves the actual underlying problem
• The Expansionist — looks for the upside and the bigger opportunity you are not seeing
• The Outsider — has zero context on purpose, so it catches the obvious thing experts skip past
• The Executor — skips the theory and tells you the single next move to make

Close the instructions with the part that does the real work: after the five have spoken, Claude weighs them against each other, drops the weak arguments, and hands you one clear verdict. Add a final line telling it to say so plainly when it is unsure rather than guessing — that single sentence is what keeps the whole thing honest.

Why it beats just asking

A normal question gets you one answer shaped by how you asked it. Forcing five named roles makes Claude argue with itself before it commits, and the disagreement is where the value lives — the Contrarian surfaces the risk your excitement hid, the Outsider names the obvious flaw, and the Executor stops it all turning into philosophy. The instruction to admit uncertainty matters just as much: a council that confidently invents an answer is worse than no council at all.

Five prompts to put it to work

These are the prompts I reach for most. Feed them the decision you are least comfortable examining first.

• Am I deluded? — Tell the Council you have already decided on a move, paste your reasoning, and ask it to separate what is solid from what is you talking yourself into it, ending with a straight yes or no
• Find my pattern — Give it your three biggest decisions from the last few years and how each turned out, and ask it to name the bias that keeps repeating and predict the next mistake it will cause
• Read my excuse — Name the thing you have been putting off and your reason for it, and ask whether that reason is a real obstacle or a dressed-up fear; if it is fear, ask for the first move to make tonight
• Attack the deal — Paste the offer, contract, or purchase you are about to accept and ask the Council what is overpriced, what is missing, what you would regret in six months, and what to negotiate first
• Stress test — Lay out a plan you are proud of, then ask each advisor to assume it failed six months from now, name the most likely cause of death, and give you the fix, ranked by what to patch first

The point of the Council is not to make Claude smarter. It is to stop it from agreeing with you — which, when you are about to make a big decision, is the most useful thing an advisor can do.Md Raihan Hasan

Build the project once and it is there every time you need a second, third, and fourth opinion. I keep mine pinned and open it whenever a decision feels too comfortable — usually the surest sign I have not actually pressure-tested it yet.