mental-models

name	mental-models
description	Apply Charlie Munger's latticework of 98 mental models to any problem. Use when user requests decision analysis, says "help me think", "apply mental model", mentions model names (inversion, bottlenecks, second-order thinking), or needs structured thinking frameworks.

Mental Models

Apply 98 cognitive frameworks from multiple disciplines to analyze problems, make decisions, and think more clearly.

This skill is backed by the mental-models CLI — the canonical interface. The CLI does selection, lookup, and structured application in one command. No file walking, no hand-parsing. Prefer the CLI.

When to Activate

• User names a specific model ("apply inversion", "use bottlenecks")
• User asks "help me think through X" or "what model fits X"
• User requests decision analysis, trade-off evaluation, or structured reasoning
• User describes a complex/ambiguous problem and wants a framework

Preflight: is the CLI available?

Run once per session via OpenClaw's exec tool:

mental-models doctor --json

{"ok": true, ...} → use the CLI workflow below.

Command not found → try uvx mental-models doctor --json (runs from PyPI without install — requires uv on PATH). If that also fails, tell the user:

Install with: pip install mental-models (or uv tool install mental-models).

CLI Workflow

Step 1 — Select models for the problem

mental-models select "<paraphrased problem>" -k 5 --json

Returns a JSON object with a models array. Each entry has slug, name, category, description, keywords, path. Pick 2–3 that best fit — prefer cross-category coverage (that's the latticework).

Step 2 — Get structured guidance for each chosen model

mental-models apply <slug> --problem "<user's problem>" --json

Returns:

• description — what the model is
• thinking_steps — the sequential framework (walk verbatim, don't paraphrase)
• coaching_questions — prompts to deepen the analysis
• when_to_avoid — failure modes (always check and surface if relevant)

Step 3 — Synthesize

• Walk each model's thinking_steps against the user's facts
• Show where the models agree, where they disagree
• End with 3–5 concrete, actionable next steps
• Name any "when to avoid" conditions that apply to this case

Other useful commands

mental-models get <slug>                 # full markdown
mental-models get <slug> --field keywords
mental-models list --category "Human Nature"
mental-models categories
mental-models which                      # resolved data path

All commands support --json. Exit codes: 0 ok, 2 not found, 3 bad args.

Discovery Heuristics (bias your select query)

• Risk / uncertainty / reversibility → inversion, probabilistic thinking, margin of safety
• Stuck / can't see options → first principles, second-order thinking, reframing
• Conflict / negotiation / competition → incentives, asymmetric warfare, trade-offs
• Complex system / unintended effects → feedback loops, emergence, bottlenecks, leverage
• Performance / optimization → bottlenecks, diminishing returns, efficiency
• People / team / behavior → incentives, social proof, biases
• Communication / persuasion → framing, audience, contrast

Core Guidelines

1. Max 3 models per analysis — quality over quantity
2. Follow thinking_steps verbatim — don't paraphrase the framework away
3. Always check when_to_avoid — warn the user if the model misfits
4. Latticework: show how chosen models connect and where they disagree
5. Be actionable: end with concrete next steps, not theory

Install the CLI

# one-shot (recommended)
uvx mental-models select "your question"

# or install globally
pip install mental-models
# or
uv tool install mental-models

Source: https://github.com/cyperx84/claude-skills-mental-models