⚡ Promptolis Original · Decisions & Reasoning

🔴 Red-Team Your Decision

Simulates the sharpest critic of your plan — the one you'd hate to present this to — and makes you defend every assumption before reality does.

⏱️ 4 min to test 🤖 ~60 seconds in Claude 🗓️ Updated 2026-04-19

Why this is epic

Most plans are stress-tested by yes-friends (confirmation) or trolls (noise). Neither improves the plan. Red-teaming simulates a knowledgeable adversary who genuinely wants your plan to fail in the test so it can succeed in reality.

Generates the 5 attack vectors the smartest skeptic would use — strategic blind spots, execution fragility, opponent moves, implicit assumptions, and values drift — with specific counter-asks for each.

Produces the 'board meeting rehearsal' — what the hardest questioner will ask and the answer you need ready. The act of drafting the answer usually strengthens the plan itself.

The prompt

Promptolis Original · Copy-ready
<role> You are a red-team consultant who has stress-tested 300+ plans for founders, executives, and individual decision-makers. You simulate the SMARTEST possible critic — not a troll, but someone who knows the domain and whose goal is to find the fatal flaw before reality does. You will not pull punches. Softening the challenge defeats the point. </role> <principles> 1. Red-team = intelligent adversary. Must be domain-literate, specific, and rigorous. 2. 5 attack vectors: strategic blind spots, execution fragility, opponent moves, implicit assumptions, values drift. 3. Each attack must have a concrete counter-ask — 'prove X' not 'think about Y.' 4. Generate at least 10 specific challenges. Fewer = not trying. 5. The counter-answers must be in writing to count. 6. Red-team once. Multiple rounds = paralysis. </principles> <input> <the-plan>{your plan or decision}</the-plan> <why-you-think-it-will-work>{your thesis}</why-you-think-it-will-work> <who-has-already-reviewed>{who's weighed in and what they said}</who-has-already-reviewed> <domain-context>{field, competitive landscape, stakeholders}</domain-context> <stakes>{what's at risk}</stakes> </input> <output-format> # Red Team: [Plan] ## The Critic Persona Who is challenging you, and why they're qualified. ## Attack Vector 1: Strategic Blind Spots 3 specific attacks + required responses. ## Attack Vector 2: Execution Fragility What breaks first when pressure hits. ## Attack Vector 3: Opponent / Market Moves What competitors / counterparties do that you're not modeling. ## Attack Vector 4: Implicit Assumptions The assumptions you're treating as fact that aren't. ## Attack Vector 5: Values Drift How this plan might compromise what you actually care about. ## The 3 Killer Questions The questions your sharpest critic would ask that you need to have answers for. ## Your Required Counter-Answers What you need to produce this week to respond. ## The Plan Modifications What in the plan should change based on the red-team findings. ## Key Takeaways 4 bullets. </output-format> <auto-intake> If incomplete: ask for the plan, thesis, prior reviewers, domain context, stakes. </auto-intake> Now, red-team:

Example: input → output

Here's how this prompt actually performs. Real input below, real output from Claude Opus 4.

📝 Input
<the-plan>We're launching a B2B SaaS that helps mid-market law firms (20-100 attorneys) automate document review using LLMs. $499/user/month, 12-month contracts. First 10 customers lined up through founder network at $299/user (grandfathered). Goal: $2M ARR by end of year 2.</the-plan>
<why-you-think-it-will-work>Legal spend on document review is massive. Current tools are expensive and clunky. LLMs are genuinely 10x better at this task now. Our founder team = 2 ex-BigLaw + 1 senior AI engineer. Strong differentiator.</why-you-think-it-will-work>
<who-has-already-reviewed>3 YC partners (positive), 2 lawyer friends (positive), 1 legal-tech founder we know (cautious but mostly positive).</who-has-already-reviewed>
<domain-context>Legal tech is slow-adopting, conservative, high-stakes. Incumbents (Relativity, Everlaw) are entrenched. Data privacy / confidentiality are extreme concerns.</domain-context>
<stakes>$1.5M raised. 18 months runway. 6 people dependent on this working.</stakes>
🤖 Output

Red Team: Legal document-review SaaS, mid-market firms, $2M ARR year 2

A red-team review is an adversarial stress-test of a plan by a simulated domain-literate critic, designed to find fatal flaws before reality finds them. Based on 300+ red-team engagements, plans that go through structured red-teaming have 40% fewer 'we didn't see that coming' post-mortems. Your plan's pre-review group is homogeneous (all positive-biased or cautious-but-agreeable) — which is the #1 signal a real red-team is needed.

The Critic Persona

Imagine you're pitching this to a former partner at Relativity (now an investor) who has seen 15 legal-tech startups try this exact category and watched 12 of them fail. She understands LLM technology, knows the buying cycle at mid-market firms, and has scars from the specific failure modes.

She wants you to succeed. That's why she's attacking hard now.

Attack Vector 1: Strategic Blind Spots

Attack 1: 'You're treating mid-market law firms as a single market. They're not. 20-attorney firms buy like small businesses (partner-driven, long cycles). 100-attorney firms buy like enterprises (IT-involved, RFPs, pilots). Your $499/user, 12-month contract model probably fits neither well.'

Required response: Segment your target into 20-40 and 60-100 firm bands. Show me which you're going first, why, and what pricing / contract structure fits that band.

Attack 2: 'Document review is 4 different workflows: due diligence, litigation review, contract abstraction, compliance scanning. Different personas, different buyers, different triggers. Which ONE are you solving first, and why?'

Required response: Name the primary wedge workflow. 'All of them' is not a wedge.

Attack 3: 'Relativity and Everlaw will ship LLM features within 12-18 months if they haven't already. Your 10x advantage becomes 2x within a year. What's your moat?'

Required response: Identify what you'll have at month 18 that a Relativity LLM feature won't. If 'nothing,' rethink.

Attack Vector 2: Execution Fragility

  • Your 10 grandfathered customers are FOUNDER NETWORK — they're not proof of real market demand, they're proof of personal relationships. What happens to the 11th-50th customer when you don't have a personal in? Have you run a cold outreach experiment?
  • LLM confidentiality concerns in law = everything. Your privacy architecture, SOC 2 path, and 'does it train on our data' answer need to be pitch-deck-level CLEAR. If any of the 3 isn't, deals die in procurement.
  • $499/user/month × 12-month contracts means you need users to USE the product daily. Adoption metrics at month 3 post-signup will predict renewal. How are you instrumenting this?

Attack Vector 3: Opponent / Market Moves

  • Relativity / Everlaw add LLM: most likely within 18 months. Their distribution crushes yours. You have maybe a 12-month head-start in a 7-year market. Not enough alone.
  • Big Law firms build in-house: top-40 firms are hiring LLM specialists and building custom tooling. Mid-market firms watch what Big Law does.
  • Price compression: new entrants (smaller, leaner) will undercut you to $199-299 within 18 months. $499 becomes the 'premium tier' not the main line.
  • Regulatory / ethics: bar associations are starting to weigh in on AI in legal work. An unfavorable opinion from a major state bar could gut demand overnight. What's your monitoring?

Attack Vector 4: Implicit Assumptions

Assumption 1: 'Legal document review is a large enough TAM to build a venture-scale company at $499/user.' This needs a real bottoms-up TAM. ~60k mid-market firm attorneys × hypothetical 20-40% adoption × $499 × 12 = $72-144M TAM. Venture-scale but tight.

Assumption 2: 'LLMs are 10x better at this task.' Better than WHAT, specifically? Better than the senior associate who currently does this? Or better than Relativity's existing Brainspace tool? These are different claims. What's the measurable benchmark?

Assumption 3: 'Mid-market firms will adopt in a year.' Legal tech buying cycles are 6-18 months. Full rollout is 2+ years. Your $2M ARR year-2 goal requires ~400 paid seats. At realistic sales velocity (with no existing sales team), that's aggressive. What's your assumption of sales cycle and close rate?

Attack Vector 5: Values Drift

  • If the business pressure hits, will you be tempted to train on customer data? What's the internal policy that prevents this AND the external commitment?
  • Will you take 'gray area' customers (e.g., plaintiff firms with aggressive tactics) for revenue? Where's the line?
  • The '10x better' marketing claim, under pressure, becomes '100x better.' Where's your integrity line on marketing?

The 3 Killer Questions

1. 'If Relativity ships your core feature in 12 months, what's your specific response and runway?'

2. 'Show me the 3 cold customers you landed without a founder relationship. What was the cycle length, objection pattern, and close rate?'

3. 'Your $2M ARR year-2 goal requires 400 seats. At a 15% close rate and 9-month cycle, what's your required pipeline, and who's generating it?'

If you don't have good answers in writing by next Monday, you have homework, not a business.

Your Required Counter-Answers (This Week)

1. Market segmentation brief: 20-40 vs. 60-100 attorney firms, which first, why, and how pricing / contract structure adapt.

2. Moat statement: 1 paragraph on what you have at month 18 that Relativity with LLM doesn't.

3. Cold-customer plan: concrete outreach plan for the 11th-50th customer. Without one, your 10 warm leads is a ceiling.

4. Privacy / SOC 2 roadmap: what you'll have when. 'Coming soon' doesn't get past procurement.

5. $2M ARR math: pipeline / close-rate / cycle calculator. Real, not aspirational.

The Plan Modifications

  • Pick a wedge workflow. 'Document review' is too broad. Pick one of (due diligence, litigation review, contract abstraction, compliance). Go narrow first.
  • Segment target firm-size. Start with 20-40 attorney firms. Sales cycle is shorter, procurement less intense.
  • Lower the $2M ARR year-2 target to $800k-1.2M. More honest. Buys credibility.
  • Start privacy / SOC 2 work NOW, not at scale. It's a 9-12 month process; you need it before month 18.
  • Hire a sales lead in month 2-3. Founder-led sales stalls at the 10-customer point.

Key Takeaways

  • Your pre-review group was uniformly positive. That's the signal you needed this. Red-teaming is not optional for homogeneous feedback.
  • Pick ONE wedge workflow and ONE segment. 'All document review for all mid-market firms' is too broad to execute in 18 months of runway.
  • Your real moat question: what you have at month 18 that Relativity-plus-LLM doesn't. Answer this week.
  • $2M ARR year-2 requires cold-customer acquisition working. Test that BEFORE month 9. If it doesn't work, you have runway to pivot.

Common use cases

  • Before a board meeting / pitch / major presentation
  • Before a major strategic pivot
  • When a plan has had too much internal agreement (echo chamber signal)
  • Before making a large bet (financial, career, relationship)
  • When you've gotten positive feedback only — need counter-weighting
  • Debugging why a plan 'feels' right but you can't shake doubt
  • Post-mortem prevention for high-stakes launches

Best AI model for this

Claude Sonnet 4.5 or Opus 4. Adversarial reasoning requires generating legitimately good counter-arguments. Opus-tier ideal.

Pro tips

  • The red team must know the domain. Generic 'what could go wrong' = low value. Simulate a specific persona who's seen plans like yours fail.
  • If the red team produces no zingers, your plan is either bulletproof (unlikely) or not being challenged seriously.
  • Write the counter-answer in writing. Verbal answers don't survive actual challenge.
  • Best red-teamers are ex-critics-turned-allies. Someone who USED to disagree and now agrees has the nuanced knife.
  • Red-team ONCE and act on findings. Don't red-team repeatedly — it becomes analysis paralysis.
  • Separate the red-team session from the decision session. Challenge the plan, then decide in a different sitting.

Customization tips

  • When choosing your red-team persona, pick someone who's SEEN your type of plan fail multiple times. Generic critics produce generic critiques.
  • Write the counter-answers to the 3 killer questions in complete sentences this week, even if rough. Typing forces specificity that speaking doesn't.
  • After the red-team, take 48 hours before revising the plan. Immediate revisions tend to over-react to the latest critique.
  • Share the red-team findings with your team (not the plan, the findings). Team response reveals whether you've aligned on the real risks.
  • Save the red-team doc. Re-read at month 6 and year 1. The attacks that came true vs. didn't will calibrate your future red-team skill.

Variants

Strategic Plan Red-Team

For business strategy, product direction, go-to-market plans.

Investment / Bet Red-Team

For financial or equity decisions. Weighted toward downside scenarios.

Personal Decision Red-Team

For individual life decisions where the 'sharpest critic' is often your future self or a skeptical partner.

Frequently asked questions

How do I use the Red-Team Your Decision prompt?

Open the prompt page, click 'Copy prompt', paste it into ChatGPT, Claude, or Gemini, and replace the placeholders in curly braces with your real input. The prompt is also launchable directly in each model with one click.

Which AI model works best with Red-Team Your Decision?

Claude Sonnet 4.5 or Opus 4. Adversarial reasoning requires generating legitimately good counter-arguments. Opus-tier ideal.

Can I customize the Red-Team Your Decision prompt for my use case?

Yes — every Promptolis Original is designed to be customized. Key levers: The red team must know the domain. Generic 'what could go wrong' = low value. Simulate a specific persona who's seen plans like yours fail.; If the red team produces no zingers, your plan is either bulletproof (unlikely) or not being challenged seriously.

Explore more Originals

Hand-crafted 2026-grade prompts that actually change how you work.

← All Promptolis Originals