# Mental Models

> Mental Models

- Skill: `sethdford/mental-models` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add sethdford/mental-models`
- Raw SKILL.md: https://api.skillmd.com/api/skills/sethdford/mental-models/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: sethdford (https://skillmd.com/u/sethdford)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/sethdford/mental-models

---

# Mental Models

Apply a curated library of lenses to frame problems, surface hidden assumptions, and reduce blind spots. Your job is **fit**, not textbook definitions: pick models that change what the user should notice or do.

## When to Use
- Ambiguous decisions, conflicting advice, or “something feels off” without a crisp diagnosis.
- Planning under uncertainty, prioritization, or evaluating other people’s incentives and constraints.

## Behaviors
Keep this reference set handy: Occam’s razor; Hanlon’s razor; circle of competence; map ≠ territory; inversion; second-order thinking; Pareto (80/20); reversibility; opportunity cost; regression to the mean; survivorship bias; compounding; margin of safety; Lindy effect; via negativa; antifragility; falsification; Bayesian updating; comparative advantage; local vs global optima.

For each situation, name **2–3** models that most shift the analysis. Briefly say **why they apply** and **what they imply** (what to check, measure, or try). If two models conflict, say so and propose how to resolve (evidence, scope, time horizon). Prefer actionable checks over naming contests.

## Examples
**Example 1:** A team blames malice for outages → apply Hanlon’s razor + map≠territory: assume coordination/tooling gaps first; verify with logs and process, not intent.

**Example 2:** “This winner will keep winning” → survivorship bias + regression to the mean: compare to full cohort; separate luck from repeatable drivers before scaling bets.