Uncertainty - The Skill I Built to Stop Lying to Myself About Risk
Updated: Aug 17
TL;DR / Key Takeaways
Core Mechanism: Leverages Stanford GSB pre-mortem and pre-celebration techniques to analyze prospective failure or unchosen success.
Behavioral Triage: Evaluates user language cues (e.g., energy level, risk dismissiveness) to select the right exercise automatically rather than asking the user directly.
Stealth Bias Fishing: Tracks cognitive biases (overconfidence, sunk cost, groupthink) implicitly and labels them only after the user recognizes the evidence.
Temporal Flexibility: Operates across the full lifecycle: pre-decision, early post-launch, or mid-project to catch emerging risks before failure occurs.
Explained like I am 5
Imagine it already happened! 🔮 Sometimes we pretend the plan already failed — "oh no, what went wrong?" and sometimes we pretend it already worked great "yay, what made it work?" Both help us plan better!
I can tell how you're feeling without asking! 👀 If you sound super excited and sure of yourself, I might pick one game. If you sound like you're not worried about anything going wrong, I might pick a different game — all without asking "which one do you want?"
I spot sneaky thinking traps... but I don't tell you right away! 🕵️ Sometimes people think "I'm sure I'm right!" even when they're not, or "I already spent so much, I can't stop now!" I notice these sneaky traps quietly, and only point them out AFTER you spot the clues yourself — like a treasure hunt!
I can help anytime — not just at the start! ⏰ Before you decide something, right after you start, or even in the middle I can jump in anytime to help catch problems before they get big!
This article describes the core principles behind a Claude Skill that I find helpful in uncertain situations.
The principles behind the skill
Every plan I've ever been excited about felt inevitable right up until it wasn't. That's not a coincidence. Confidence and clear-sightedness don't move together, but they trade off. The more invested I am in a path, the worse I am at seeing what could break it. So the skill I use most before committing to anything real isn't a planning tool. It's an interrogation tool. I call it the Uncertainty Navigator, and it's built on Paul Pfleiderer's pre-mortem and pre-celebration method from Stanford GSB.
Pre-mortem: assume the plan already failed completely, then work backward to explain why.
Pre-celebration: assume the option you rejected became a massive success, then explain why that happened.
Both use prospective hindsight, standing in an imagined future and looking back, which activates different thinking than "what could go wrong" asked in the present tense.
Here's a snippet from how the skill decides which exercise to run:
"Deploy Pre-Mortem if you observe: Heavy investment in describing the favored option (detail, energy, certainty). Alternatives are barely mentioned or only appear when prompted. Risks are framed as 'unlikely', 'handled', or glossed over."I wrote it this way because asking someone "are you overconfident?" gets you a defensive no, every time. Nobody self-reports their own blind spot on request. So the skill doesn't ask. It listens for the pattern instead, how much energy goes into defending the favored option versus how quickly the alternatives get waved off and picks the exercise based on behavior, not self-assessment.
The second piece worth showing is the bias-naming sequence:
"Do NOT name these as biases yet. Take mental note of patterns. Return to them at Step 4... Name biases late. The goal of bias fishing is insight, not confrontation. If a user gets defensive about being told they're overconfident, you've lost the session."This is the part people usually get wrong when they try to build something like this. The instinct is to diagnose fast ("you're anchoring," "that's sunk cost") because naming the bias feels like the value delivered.
Naming a bias before someone has felt the evidence for themselves just produces resistance. So the skill collects three to five data points quietly, runs the actual exercise, and only reflects the pattern back once there's enough material for the person to recognize it themselves rather than be told it. Confirmation happens before the label does.
The last snippet is the one I think matters most, because it's the part that keeps this from becoming a one-time trick:
"Temporal Flexibility: The pre-mortem is not only a pre-decision tool. It can be applied mid-project or post-decision... to stress-test the current trajectory, surface emerging risks, and trigger course corrections before failure actually happens."Most premortem advice treats it as a pre-decision ritual: do it once, before you commit, then move on. The plans that actually fail rarely fail because of something unknowable at the start, they fail because of something that became visible at month six and got ignored because nobody wanted to reopen the question.
So the skill is built to run again, mid-flight, with the frame shifted from "it's failed" to "it's off-track and you're doing an honest review." Same mechanism, different entry point. The goal isn't a decision gate you pass once, it's a habit of asking, at intervals, what you're currently choosing not to look at.
None of this replaces judgment. A pre-mortem doesn't tell you what to do, it tells you what you're not currently seeing. The value is entirely in how honestly you let the imagined failure land before you rush to defend against it.
The anatomy of the skill
Metadata & Behavioral Triggers
Activates explicitly for pre-mortems/stress tests, or implicitly when a user displays unprompted overconfidence, dismissiveness, or tunnel vision.
Core Methodology: Prospective Hindsight
Forces the user to stand in an imagined future looking backward, bypassing present-tense defensive reasoning.
"Prospective hindsight is the engine. Keep returning the user to 'You're in the future, looking back.' This activates different thinking than 'What might go wrong?'"
Step 0: Stealth Behavioral Triage
Bypasses direct self-assessment questions, measuring speech patterns and option framing to select the appropriate exercise.
"Deploy Pre-Mortem if you observe: Heavy investment in describing the favored option (detail, energy, certainty). Alternatives are barely mentioned or only appear when prompted. Risks are framed as 'unlikely', 'handled', or glossed over."
Steps 1 & 4: Indirect Bias Fishing & Delayed Reflection
Gathers behavioral data silently through probing questions, runs the exercises, and introduces bias labels late in the process.
"Do NOT name these as biases yet. Take mental note of patterns. Return to them at Step 4... Name biases late. The goal of bias fishing is insight, not confrontation. If a user gets defensive about being told they're overconfident, you've lost the session."
Steps 2 & 3: Dual-Vector Stress Testing
Pre-Mortem: Assumes total failure of the primary choice to probe systemic, human, and environmental failure modes.
Pre-Celebration: Assumes massive success of a rejected alternative to uncover overlooked upside.
"Protect the pre-celebration. Users often resist imagining success for alternatives they've rejected. Hold the frame."
Lifecycle Adaptation & Temporal Variants
Extends utility beyond initial pre-decision gates into ongoing mid-project trajectory checks.
"The pre-mortem is not only a pre-decision tool. It can be applied mid-project or post-decision... to stress-test the current trajectory, surface emerging risks, and trigger course corrections before failure actually happens."
Step 5: High-Density Synthesis
Converts qualitative reflection into actionable risk controls, ranking the top failure modes by probability $\times$ impact and assigning immediate mitigation steps.
Want to put this + and the rest of the decision-making skills to work on something you're actually facing? I am giving away the skills for an one-hour discovery interview. Sign up here.
