zhongwei/gh-lyndonkl-claude
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Initial commit

Browse Source
This commit is contained in:
Zhongwei Li
2025-11-30 08:38:26 +08:00
commit 41d9f6b189
304 changed files with 98322 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

462
skills/forecast-premortem/SKILL.md Normal file
View File

@@ -0,0 +1,462 @@
---
name: forecast-premortem
description: Use to stress-test predictions by assuming they failed and working backward to identify why. Invoke when confidence is high (>80% or <20%), need to identify tail risks and unknown unknowns, or want to widen overconfident intervals. Use when user mentions premortem, backcasting, what could go wrong, stress test, or black swans.
---
# Forecast Pre-Mortem
## Table of Contents
- [What is a Forecast Pre-Mortem?](#what-is-a-forecast-pre-mortem)
- [When to Use This Skill](#when-to-use-this-skill)
- [Interactive Menu](#interactive-menu)
- [Quick Reference](#quick-reference)
- [Resource Files](#resource-files)
---
## What is a Forecast Pre-Mortem?
A **forecast pre-mortem** is a stress-testing technique where you assume your prediction has already failed and work backward to construct the history of how it failed. This reveals blind spots, tail risks, and overconfidence.
**Core Principle:** Invert the problem. Don't ask "Will this succeed?" Ask "It has failed - why?"
**Why It Matters:**
- Defeats overconfidence by forcing you to imagine failure
- Identifies specific failure modes you hadn't considered
- Transforms vague doubt into concrete risk variables
- Widens confidence intervals appropriately
- Surfaces "unknown unknowns"
**Origin:** Gary Klein's "premortem" technique, adapted for probabilistic forecasting
---
## When to Use This Skill
Use this skill when:
- **High confidence** (>80% or <20%) - Most likely to be overconfident
- **Feeling certain** - Certainty is a red flag in forecasting
- **Prediction is important** - Stakes are high, need robustness
- **After inside view analysis** - Used specific details, might have missed big picture
- **Before finalizing forecast** - Last check before committing
Do NOT use when:
- Confidence already low (~50%) - You're already uncertain
- Trivial low-stakes prediction - Not worth the time
- Pure base rate forecasting - Premortem is for inside view adjustments
---
## Interactive Menu
**What would you like to do?**
### Core Workflows
**1. [Run a Failure Premortem](#1-run-a-failure-premortem)** - Assume prediction failed, explain why
**2. [Run a Success Premortem](#2-run-a-success-premortem)** - For pessimistic predictions (<20%)
**3. [Dragonfly Eye Perspective](#3-dragonfly-eye-perspective)** - View failure through multiple lenses
**4. [Identify Tail Risks](#4-identify-tail-risks)** - Find black swans and unknown unknowns
**5. [Adjust Confidence Intervals](#5-adjust-confidence-intervals)** - Quantify the adjustment
**6. [Learn the Framework](#6-learn-the-framework)** - Deep dive into methodology
**7. Exit** - Return to main forecasting workflow
---
## 1. Run a Failure Premortem
**Let's stress-test your prediction by imagining it has failed.**
```
Failure Premortem Progress:
- [ ] Step 1: State the prediction and current confidence
- [ ] Step 2: Time travel to failure
- [ ] Step 3: Write the history of failure
- [ ] Step 4: Identify concrete failure modes
- [ ] Step 5: Assess plausibility and adjust
```
### Step 1: State the prediction and current confidence
**Tell me:**
1. What are you predicting?
2. What's your current probability?
3. What's your confidence interval?
**Example:** "This startup will reach $10M ARR within 2 years" - Probability: 75%, CI: 60-85%
### Step 2: Time travel to failure
**The Crystal Ball Exercise:**
Jump forward to the resolution date. **It is now [resolution date]. The event did NOT happen.** This is a certainty. Do not argue with it.
**How does it feel?** Surprising? Expected? Shocking? This emotional response tells you about your true confidence.
### Step 3: Write the history of failure
**Backcasting Narrative:** Starting from the failure point, work backward in time. Write the story of how we got here.
**Prompts:**
- "The headlines that led to this were..."
- "The first sign of trouble was when..."
- "In retrospect, we should have known because..."
- "The critical mistake was..."
**Frameworks to consider:**
- **Internal friction:** Team burned out, co-founders fought, execution failed
- **External shocks:** Regulation changed, competitor launched, market shifted
- **Structural flaws:** Unit economics didn't work, market too small, tech didn't scale
- **Black swans:** Pandemic, war, financial crisis, unexpected disruption
See [Failure Mode Taxonomy](resources/failure-mode-taxonomy.md) for comprehensive categories.
### Step 4: Identify concrete failure modes
**Extract specific, actionable failure causes from your narrative.**
For each failure mode: (1) What happened, (2) Why it caused failure, (3) How likely it is, (4) Early warning signals
**Example:**
| Failure Mode | Mechanism | Likelihood | Warning Signals |
|--------------|-----------|------------|-----------------|
| Key engineer quit | Lost technical leadership, delayed product | 15% | Declining code commits, complaints |
| Competitor launched free tier | Destroyed unit economics | 20% | Hiring spree, beta leaks |
| Regulation passed | Made business model illegal | 5% | Proposed legislation, lobbying |
### Step 5: Assess plausibility and adjust
**The Plausibility Test:**
Ask yourself:
- **How easy was it to write the failure narrative?**
- Very easy → Drop confidence by 15-30%
- Very hard, felt absurd → Confidence was appropriate
- **How many plausible failure modes did you identify?**
- 5+ modes each >5% likely → Too much uncertainty for high confidence
- 1-2 modes, low likelihood → Confidence can stay high
- **Did you discover any "unknown unknowns"?**
- Yes, multiple → Widen confidence intervals by 20%
- No, all known risks → Confidence appropriate
**Quantitative Method:** Sum the probabilities of failure modes:
```
P(failure) = P(mode_1) + P(mode_2) + ... + P(mode_n)
```
If this sum is greater than `1 - your_current_probability`, your probability is too high.
**Example:** Current success: 75% (implied failure: 25%), Sum of failure modes: 40%
**Conclusion:** Underestimating failure risk by 15%, **Adjusted:** 60% success
**Next:** Return to [menu](#interactive-menu) or document findings
---
## 2. Run a Success Premortem
**For pessimistic predictions - assume the unlikely success happened.**
```
Success Premortem Progress:
- [ ] Step 1: State pessimistic prediction (<20%)
- [ ] Step 2: Time travel to success
- [ ] Step 3: Write the history of success
- [ ] Step 4: Identify how you could be wrong
- [ ] Step 5: Assess and adjust upward if needed
```
### Step 1: State pessimistic prediction
**Tell me:** (1) What low-probability event are you predicting? (2) Why is your confidence so low?
**Example:** "Fusion energy will be commercialized by 2030" - Probability: 10%, Reasoning: Technical challenges too great
### Step 2: Time travel to success
**It is now 2030. Fusion energy is commercially available.** This happened. It's real. How?
### Step 3: Write the history of success
**Backcasting the unlikely:** What had to happen for this to occur?
- "The breakthrough came when..."
- "We were wrong about [assumption] because..."
- "The key enabler was..."
- "In retrospect, we underestimated..."
### Step 4: Identify how you could be wrong
**Challenge your pessimism:**
- Are you anchoring too heavily on current constraints?
- Are you underestimating exponential progress?
- Are you ignoring parallel approaches?
- Are you biased by past failures?
### Step 5: Assess and adjust upward if needed
If success narrative was surprisingly plausible, increase probability.
**Next:** Return to [menu](#interactive-menu)
---
## 3. Dragonfly Eye Perspective
**View the failure through multiple conflicting perspectives.**
The dragonfly has compound eyes that see from many angles simultaneously. We simulate this by adopting radically different viewpoints.
```
Dragonfly Eye Progress:
- [ ] Step 1: The Skeptic (why this will definitely fail)
- [ ] Step 2: The Fanatic (why failure is impossible)
- [ ] Step 3: The Disinterested Observer (neutral analysis)
- [ ] Step 4: Synthesize perspectives
- [ ] Step 5: Extract robust failure modes
```
### Step 1: The Skeptic
**Channel the harshest critic.** You are a short-seller, a competitor, a pessimist. Why will this DEFINITELY fail?
**Be extreme:** Assume worst case, highlight every flaw, no charity, no benefit of doubt
**Output:** List of failure reasons from skeptical view
### Step 2: The Fanatic
**Channel the strongest believer.** You are the founder's mother, a zealot, an optimist. Why is failure IMPOSSIBLE?
**Be extreme:** Assume best case, highlight every strength, maximum charity and optimism
**Output:** List of success reasons from optimistic view
### Step 3: The Disinterested Observer
**Channel a neutral analyst.** You have no stake in the outcome. You're running a simulation, analyzing data dispassionately.
**Be analytical:** No emotional investment, pure statistical reasoning, reference class thinking
**Output:** Balanced probability estimate with reasoning
### Step 4: Synthesize perspectives
**Find the overlap:** Which failure modes appeared in ALL THREE perspectives?
- Skeptic mentioned it
- Even fanatic couldn't dismiss it
- Observer identified it statistically
**These are your robust failure modes** - the ones most likely to actually happen.
### Step 5: Extract robust failure modes
**The synthesis:**
| Failure Mode | Skeptic | Fanatic | Observer | Robust? |
|--------------|---------|---------|----------|---------|
| Market too small | Definitely | Debatable | Base rate suggests yes | YES |
| Execution risk | Definitely | No way | 50/50 | Maybe |
| Tech won't scale | Definitely | Already solved | Unknown | Investigate |
Focus adjustment on the **robust** failures that survived all perspectives.
**Next:** Return to [menu](#interactive-menu)
---
## 4. Identify Tail Risks
**Find the black swans and unknown unknowns.**
```
Tail Risk Identification Progress:
- [ ] Step 1: Define what counts as "tail risk"
- [ ] Step 2: Systematic enumeration
- [ ] Step 3: Impact × Probability matrix
- [ ] Step 4: Set kill criteria
- [ ] Step 5: Monitor signposts
```
### Step 1: Define what counts as "tail risk"
**Criteria:** Low probability (<5%), High impact (would completely change outcome), Outside normal planning, Often exogenous shocks
**Examples:** Pandemic, war, financial crisis, regulatory ban, key person death, natural disaster, technological disruption
### Step 2: Systematic enumeration
**Use the PESTLE framework for comprehensive coverage:**
- **Political:** Elections, coups, policy changes, geopolitical shifts
- **Economic:** Recession, inflation, currency crisis, market crash
- **Social:** Cultural shifts, demographic changes, social movements
- **Technological:** Breakthrough inventions, disruptions, cyber attacks
- **Legal:** New regulations, lawsuits, IP challenges, compliance changes
- **Environmental:** Climate events, pandemics, natural disasters
For each category, ask: "What low-probability event would kill this prediction?"
See [Failure Mode Taxonomy](resources/failure-mode-taxonomy.md) for detailed categories.
### Step 3: Impact × Probability matrix
**Plot your tail risks:**
```
High Impact
│
│ [Pandemic] [Key Founder Dies]
│
│
│ [Recession] [Competitor Emerges]
│
└─────────────────────────────────────→ Probability
Low High
```
**Focus on:** High impact, even if very low probability
### Step 4: Set kill criteria
**For each major tail risk, define the "kill criterion":**
**Format:** "If [event X] happens, probability drops to [Y]%"
**Examples:**
- "If FDA rejects our drug, probability drops to 5%"
- "If key engineer quits, probability drops to 30%"
- "If competitor launches free tier, probability drops to 20%"
- "If regulation passes, probability drops to 0%"
**Why this matters:** You now have clear indicators to watch
### Step 5: Monitor signposts
**For each kill criterion, identify early warning signals:**
| Kill Criterion | Warning Signals | Check Frequency |
|----------------|----------------|-----------------|
| FDA rejection | Phase 2 trial results, FDA feedback | Monthly |
| Engineer quit | Code velocity, satisfaction surveys | Weekly |
| Competitor launch | Hiring spree, beta leaks, patents | Monthly |
| Regulation | Proposed bills, lobbying, hearings | Quarterly |
**Setup monitoring:** Calendar reminders, news alerts, automated tracking
**Next:** Return to [menu](#interactive-menu)
---
## 5. Adjust Confidence Intervals
**Quantify how much the premortem should change your bounds.**
```
Confidence Interval Adjustment Progress:
- [ ] Step 1: State current CI
- [ ] Step 2: Evaluate premortem findings
- [ ] Step 3: Calculate width adjustment
- [ ] Step 4: Set new bounds
- [ ] Step 5: Document reasoning
```
### Step 1: State current CI
**Current confidence interval:** Lower bound: __%, Upper bound: __%, Width: ___ percentage points
### Step 2: Evaluate premortem findings
**Score your premortem on these dimensions (1-5 each):**
1. **Narrative plausibility** - 1 = Failure felt absurd, 5 = Failure felt inevitable
2. **Number of failure modes** - 1 = Only 1-2 unlikely modes, 5 = 5+ plausible modes
3. **Unknown unknowns discovered** - 1 = No surprises, all known, 5 = Many blind spots revealed
4. **Dragonfly synthesis** - 1 = Perspectives diverged completely, 5 = All agreed on failure modes
**Total score:** __ / 20
### Step 3: Calculate width adjustment
**Adjustment formula:**
```
Width multiplier = 1 + (Score / 20)
```
**Examples:**
- Score = 4/20 → Multiplier = 1.2 → Widen CI by 20%
- Score = 10/20 → Multiplier = 1.5 → Widen CI by 50%
- Score = 16/20 → Multiplier = 1.8 → Widen CI by 80%
**Current width:** ___ points, **Adjusted width:** Current × Multiplier = ___ points
### Step 4: Set new bounds
**Method: Symmetric widening around current estimate**
```
New lower = Current estimate - (Adjusted width / 2)
New upper = Current estimate + (Adjusted width / 2)
```
**Example:** Current: 70%, CI: 60-80% (width = 20), Score: 12/20, Multiplier: 1.6, New width: 32, **New CI: 54-86%**
### Step 5: Document reasoning
**Record:** (1) What failure modes drove the adjustment, (2) Which perspective was most revealing, (3) What unknown unknowns were discovered, (4) What monitoring you'll do going forward
**Next:** Return to [menu](#interactive-menu)
---
## 6. Learn the Framework
**Deep dive into the methodology.**
### Resource Files
📄 **[Premortem Principles](resources/premortem-principles.md)** - Why humans are overconfident, hindsight bias and outcome bias, the power of inversion, research on premortem effectiveness
📄 **[Backcasting Method](resources/backcasting-method.md)** - Structured backcasting process, temporal reasoning techniques, causal chain construction, narrative vs quantitative backcasting
📄 **[Failure Mode Taxonomy](resources/failure-mode-taxonomy.md)** - Comprehensive failure categories, internal vs external failures, preventable vs unpreventable, PESTLE framework for tail risks, kill criteria templates
**Next:** Return to [menu](#interactive-menu)
---
## Quick Reference
### The Premortem Commandments
1. **Assume failure is certain** - Don't debate whether, debate why
2. **Be specific** - Vague risks don't help; concrete mechanisms do
3. **Use multiple perspectives** - Skeptic, fanatic, observer
4. **Quantify failure modes** - Estimate probability of each
5. **Set kill criteria** - Know what would change your mind
6. **Monitor signposts** - Track early warning signals
7. **Widen CIs** - If premortem was too easy, you're overconfident
### One-Sentence Summary
> Assume your prediction has failed, write the history of how, and use that to identify blind spots and adjust confidence.
### Integration with Other Skills
- **Before:** Use after inside view analysis (you need something to stress-test)
- **After:** Use `scout-mindset-bias-check` to validate adjustments
- **Companion:** Works with `bayesian-reasoning-calibration` for quantitative updates
- **Feeds into:** Monitoring systems and adaptive forecasting
---
## Resource Files
📁 **resources/**
- [premortem-principles.md](resources/premortem-principles.md) - Theory and research
- [backcasting-method.md](resources/backcasting-method.md) - Temporal reasoning process
- [failure-mode-taxonomy.md](resources/failure-mode-taxonomy.md) - Comprehensive failure categories
---
**Ready to start? Choose a number from the [menu](#interactive-menu) above.**