Power Prompts: Pre-Mortems, Steelmanning & Red-Teaming
Three frameworks that force rigor into decisions: pre-mortems, steelmanning, and red-teaming, with the exact prompts to run each.
You've done the research. Now you have to decide. This is where most people close Claude and go with their gut. Don't. Claude is exceptional at decision support, not because it decides for you, but because it forces rigor into your thinking before you commit.
The enemy is optimism bias: we see the upside more readily than the downside and underweight risks we haven't personally experienced. Three frameworks counter it: the pre-mortem, steelmanning, and red-teaming. Each is a different attack angle on the same goal, catching what you'd otherwise discover only after committing.
This lesson gives you the exact prompt for each. Copy it, adapt the context, run it before any significant decision.
Framework 1: The Pre-Mortem
A pre-mortem flips "what could go wrong?" into "it went wrong. What happened?" The past-tense framing bypasses optimism bias and produces specific failure modes instead of hedges.
The prompt:
"Run a pre-mortem on [decision]. Assume it's [6 months / 1 year] from now and it failed. Give me the 5 most likely failure modes, ranked by probability. For each: what went wrong, the early warning sign we should have caught, and what we could do now to reduce the risk. Don't sugarcoat it."
Claude defaults to hedged analysis; "don't sugarcoat it" gives it permission to be direct.
When: before any launch, major hire, pivot, or big spend. Then: address each of the top 3 failure modes, or consciously accept it.
Framework 2: Steelmanning
A straw man is the weakest version of an opposing argument, built to be knocked down. A steel man is the strongest version, the one you actually have to engage with. Claude can build it for the option you're not choosing.
The prompt:
"I'm leaning toward [option]. Steelman the case for [the alternative]: the evidence I'm probably underweighting, the assumptions in my preferred option that could be wrong, and the scenarios where the alternative clearly wins. Be genuinely persuasive."
"Be genuinely persuasive" is critical. Without it, Claude produces a balanced summary that never stress-tests your position.
When: after you've decided, before you commit resources. Then: if you can answer every point, commit with confidence. If you can't, address the gaps first.
Framework 3: Red-Teaming
Red-teaming comes from military and security practice: a dedicated team attacks your plan to find weaknesses before an adversary does. A pre-mortem imagines failure; steelmanning argues the other side; red-teaming attacks the execution of your specific plan, every assumption, claim, and dependency.
The prompt:
"Red-team this plan: [paste plan]. Attack it. Find every assumption that could be wrong, every dependency that could fail, every unsupported claim. Be brutal and specific, not 'this might not work' but 'this will fail because X.' For each attack, tell me the defense."
Asking for defenses turns criticism into patches.
When: before launches, pitches, negotiations. For very high-stakes decisions, run all three in sequence: pre-mortem, then steelman, then red-team. Each surfaces different risks.
Pre-mortem: assume it failed and ask what happened; past-tense framing bypasses optimism bias. Steelmanning: ask Claude for the strongest case for the option you're not choosing. Red-teaming: ask Claude to attack your specific plan and supply a defense for each vulnerability. Run them before high-stakes decisions to catch what you'd otherwise learn only after committing.