Decision Mental Models

Apply the most relevant mental models (First Principles, Inversion, Second-Order Thinking, Occam's Razor, and 16 others) to any problem or decision, surfaces non-obvious insights by explicitly matching and working through 2-3 models per query.

## Overview

This skill turns any problem or decision into a structured thinking exercise. Given a user's situation, the agent selects the 2–3 most relevant mental models from a library of 20 well-established frameworks, applies each one explicitly with worked reasoning, and surfaces non-obvious insights the user is unlikely to reach through intuitive thinking alone. The skill does not give advice, it reveals structure.

---

## When to use this skill

**Trigger on any of the following:**

- User presents a problem they are stuck on or unsure how to approach
- User faces a decision between two or more options
- User asks "how should I think about X?" or "what am I missing?"
- User describes a situation that feels complex, unclear, or emotionally loaded
- User says "help me decide", "what would you do", "I keep going in circles", or "what's the best way to approach this?"
- User is planning something and wants to stress-test assumptions
- User is explaining a failure and wants to understand what went wrong

**Do NOT trigger on:**

- Requests for factual lookups, definitions, or research summaries with no decision component
- Pure creative writing tasks
- Technical tasks (code, data, math) where reasoning frameworks would not add value
- Small talk or casual conversation with no substantive problem

---

## Instructions

### Step 1 — Extract the problem

Read the user's message carefully. Identify:
- The core tension or question (what decision or problem is actually at stake?)
- The domain (career, business, relationships, learning, finance, strategy, etc.)
- Whether the user is stuck on understanding, deciding, or planning
- Any assumptions the user appears to be making without questioning

If the problem is too vague to apply models to, ask one clarifying question before proceeding. Do not ask more than one question.

---

### Step 2 — Select 2–3 mental models

Using the **Mental Model Library** (Step 4), identify which 2–3 models are most structurally relevant to this specific problem. Use the **When to Apply** column for each model to guide your selection.

Prioritize models that:
- Reveal a blind spot the user's framing is hiding
- Challenge the default direction of their thinking
- Complement each other without being redundant

Do NOT select models just because they are famous. Select them because they fit.

---

### Step 3 — Apply each model explicitly

For each selected model, structure your output as follows:

```
### [Model Name]
**What it asks:** [The core question this model poses]
**Applied to your situation:** [Work through the model using the user's specific details]
**Non-obvious insight:** [State the one thing this model reveals that straight thinking would miss]
```

Do not summarize the model generically. Every sentence in the application must reference the user's actual situation.

---

### Step 4 — Mental Model Library

Below are the 20 mental models available to this skill. Each entry includes: what it is, when to apply it, the core question it answers, and a worked example from real life.

---

#### 1. First Principles Thinking
**What it is:** Decompose a problem down to its most fundamental verified truths, then reason upward from those truths rather than from analogy or convention.
**When to apply:** When the user seems to be accepting inherited assumptions, when conventional solutions feel expensive or broken, or when the user is trying to innovate rather than iterate.
**Core question it answers:** What do I know is actually, provably true, and what am I just assuming?
**Worked example:** Elon Musk asked why rockets cost $65M. Industry said "that's just what rockets cost." First principles: what are the raw material costs? ~2% of retail price. So he founded SpaceX to manufacture rockets from scratch, cutting costs by an order of magnitude.

---

#### 2. Inversion
**What it is:** Instead of asking "how do I achieve X?", ask "what would guarantee I fail at X?" Then systematically avoid those failure conditions.
**When to apply:** When the user is planning something ambitious and overlooking risks, when they are stuck trying to force a positive solution, or when the problem involves avoiding a bad outcome.
**Core question it answers:** What would make this go terribly wrong, and am I already doing those things?
**Worked example:** Charlie Munger didn't ask "how do I build wealth?" He asked "what behaviors reliably destroy wealth?", speculation, debt, envy, short-termism. Avoiding those behaviors consistently was itself the strategy.

---

#### 3. Second-Order Thinking
**What it is:** Go beyond the immediate consequence of a decision. Ask "and then what?" repeatedly to trace downstream effects, the effects of the effects.
**When to apply:** When the user's plan has a compelling first-order benefit, when a decision affects many people or a complex system, or when past "solutions" in this domain have created new problems.
**Core question it answers:** What happens after what happens next?
**Worked example:** A government lowers interest rates to stimulate the economy (first order: cheap credit, more spending). Second order: asset price inflation. Third order: wealth inequality widens as asset owners benefit disproportionately. The solution created a new problem.

---

#### 4. Occam's Razor
**What it is:** When multiple explanations exist, prefer the one requiring the fewest assumptions. Complexity is a liability until it is proven necessary.
**When to apply:** When the user is constructing an elaborate theory to explain something, when they are overthinking a situation, or when they are choosing between a simple and complex solution.
**Core question it answers:** Am I adding complexity that isn't justified by evidence?
**Worked example:** A startup's sales dropped 40% last month. They hypothesize: algorithm change, competitor launch, pricing mismatch, seasonal shift, and three other theories. Occam's Razor: check the simplest explanation first — did a key sales rep leave or did a major client churn? Usually one change explains the data.

---

#### 5. The Map Is Not the Territory
**What it is:** Your mental model of reality is always an abstraction. Every map omits details. Mistaking your representation of a situation for the situation itself causes error.
**When to apply:** When the user is highly confident in their understanding, when they are surprised that something "isn't working as expected," or when their plan was built from theory rather than direct observation.
**Core question it answers:** Where might my mental model be misleading me about what's actually happening?
**Worked example:** A manager designs a perfect-looking process on paper. In practice, employees skip step 3 every time because the tool required doesn't work on their machines. The map (process doc) was not the territory (how work actually gets done).

---

#### 6. Circle of Competence
**What it is:** You have domains where your knowledge is deep and reliable, and domains where it is shallow and borrowed. Decisions made inside your circle are more likely to succeed than those made outside it.
**When to apply:** When the user is entering a new domain, when confidence seems disproportionate to actual experience, or when they are relying on second-hand knowledge to make a high-stakes choice.
**Core question it answers:** Am I operating inside or outside the boundary of what I actually know well?
**Worked example:** Warren Buffett consistently declined to invest in tech companies during the dot-com boom because they were outside his circle. While others chased 100x returns and lost everything, his disciplined boundary-keeping preserved capital.

---

#### 7. Opportunity Cost
**What it is:** Every choice you make eliminates other choices. The true cost of any decision includes the value of the best alternative you gave up.
**When to apply:** When the user is evaluating an option in isolation, when they are debating whether to continue something they've already sta