Prompt · Role Prompting · Intermediate

Red Teamer: Get a Ranked List of What Could Go Wrong

Pick the attackers. Rank the damage. Fix the worst first.

This red teamer prompt attacks your plan from three seats, each with its own motive. Every flaw comes back ranked by damage, so you know where to start.

About This Prompt

Author: Adam Bloom
AI Technique: Role Prompting
Level: Intermediate
Best For: Property managers and brokers · Lease-renewal plans
Last Updated: 2026-11-11
Last Reviewed: 2026-11-11

Red Teamer Prompt Template

You are an expert in [FIELD] as a red teamer with [YEARS] years reviewing [TYPE OF BUSINESS]. You focus on how plans fail in practice. Attack my plan from three seats: [ATTACKER 1], [ATTACKER 2] and [ATTACKER 3]. Give each seat a motive. Context I’m giving you: – Business: [SIZE AND MODEL] – Constraints: [BUDGET, TIMELINE, PEOPLE] Answer in this order: 1. The strongest objection from each seat, in two sentences. 2. Every flaw, ranked high, medium or low, with one reason. 3. The one fix you would make first, and why. 4. What to check to confirm each high flaw. Rule: Don’t praise the plan, and flag any claim you can’t support with evidence. My plan: [MY PLAN]

Free to use. Copy it, change it, share it. No credit needed.

What to Fill In

[FIELD]: the area the attacker should know, such as residential property management.
[YEARS] and [TYPE OF BUSINESS]: how seasoned the attacker is, and whose plans they review.
[ATTACKER 1 to 3]: the people who could say no or lose money. Try a tenant, an owner and a rival.
[SIZE AND MODEL]: your portfolio, your team and how you earn.
[BUDGET, TIMELINE, PEOPLE]: your limits. Use real figures, or leave the line out.
[MY PLAN]: paste the plan, with numbers and dates.

Example: A Red Teamer Prompt on a Lease Renewal Plan

Illustrative example, not a client result

Input: Field: residential property management. Years: 15. Type of business: small property management firms. Attackers: a long-term tenant, an owner who wants steady cash flow, a rival landlord. Business: 24 units, two staff, 94% occupancy. Constraints: renewals due in 90 days, no budget for upgrades, one leasing agent. Plan: raise renewal rents 8% on all 24 units and offer 12-month terms only.

Prompt as sent:

You are an expert in residential property management as a red teamer with 15 years reviewing small property management firms. You focus on how plans fail in practice. Attack my plan from three seats: a long-term tenant, an owner who wants steady cash flow and a rival landlord. Give each seat a motive. Context I’m giving you: – Business: 24 units, two staff, 94% occupancy – Constraints: renewals due in 90 days, no budget for upgrades, one leasing agent Answer in this order: 1. The strongest objection from each seat, in two sentences. 2. Every flaw, ranked high, medium or low, with one reason. 3. The one fix you would make first, and why. 4. What to check to confirm each high flaw. Rule: Don’t praise the plan, and flag any claim you can’t support with evidence. My plan: Raise renewal rents 8% on all 24 units and offer 12-month terms only.

Output (shortened):

Objections. Tenant: “I’ve paid on time for years. An 8% jump with no reason feels like a penalty, so I’ll start looking.” Owner: “A vacant unit costs me more than a smaller raise earns. Why risk units that are already full?” Rival: “Every flat raise letter you send is free advertising for me.”
Ranked flaws. High: one flat raise ignores which tenants are most likely to leave, and those are the ones you can least afford to lose. High: nothing explains the raise to tenants. Medium: 12-month terms remove the flexibility some tenants want. Low: one leasing agent can’t handle a wave of move-outs.
First fix. Tier the units. Raise less for long, reliable tenants, and more where rents sit below the market. Send every notice with a reason.
Checks. Your cost per turnover: vacant days, repairs, leasing time. Rents on five comparable nearby units. How many tenants are month-to-month today.

Flag: The 8% is your figure, and every number here is invented for illustration. The AI can’t see local rents, so check them before you decide.

Why the Red Teamer Prompt Works

Three seats give three different motives. A tenant, an owner and a rival don’t worry about the same things. Ranking by damage puts the flaw that could sink the plan at the top.

When the Red Teamer Prompt Doesn’t Work

The seats are roles the AI plays. None is a real tenant, owner or rival, and none has seen your market.
In one benchmark, debate setups did not reliably beat simpler prompts. They also cost more (Smit et al., 2024).
A confident wrong objection can win. Debating models sometimes drop a right answer for a confident wrong one (Srivastava et al., 2025). Check each objection before you act.

Research Behind This Prompt

Liang et al. (2023), Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate. Supports. Two models argued opposite sides while a judge ruled. That helped on arithmetic traps and translation. It pushed models off an early wrong answer. The judge tended to favor the side running its own model.
Smit et al. (2024), Should we be going MAD? A Look at Multi-Agent Debate Strategies for LLMs. Mixed. Debate setups did not reliably beat simpler prompts like self-consistency, and they cost more. But tuning how much the debaters agree changed results a lot. The authors blame sensitivity to settings, not the idea itself.
Srivastava et al. (2025), Debate, Train, Evolve: Self-Evolution of Language Model Reasoning. Challenges. In plain debates, models often dropped a correct answer to copy a confident wrong one. They also favored longer arguments, sound or not. So don’t let a sure-sounding objection settle anything.

What the Evidence Says

Mixed. Opposing seats helped in one study. But debate setups did not reliably beat simpler prompts. And models can cave to a confident wrong answer. Use the seats to find objections, then check each one yourself.

Sources last checked 2026-11-11.

Red Teamer Prompt FAQ

Why use three seats instead of one attacker?

Different people fear different things. A tenant, an owner and a rival each find flaws the others miss.

How do I choose the seats?

Pick the people who could say no to your plan, or lose money from it. A customer, a payer and a competitor usually cover it.

Why rank flaws by damage?

A flat list hides the one flaw that could sink the plan. Ranking puts that one on top.

Why ask what to check?

An objection is only a claim. A check turns it into something you can confirm in an hour.

Are the seats real people?

No. They are roles the AI plays. None of them has seen your market or your numbers.

Want this applied to your business?

No pitch. Just a conversation.

Book a 30-minute call