Turn Claude Code into a thinking partner with 98 Munger-style mental models bundled as a single skill.
A latticework for better decisions, one prompt away.
A CLI, Claude Code skill, MCP server, and Python library — all powered by the same 98 Munger-style mental models. Ask for "inversion," "find the bottleneck," or "help me think through X" from any harness, and the same deterministic selector picks the right models, walks their Thinking Steps, and returns actionable insight.
| Interface | Install | Use |
|---|---|---|
| Claude Code skill (primary) | symlink/copy .claude/skills/mental-models/ to ~/.claude/skills/ |
auto-activates on trigger phrases |
| Portable skill for OpenClaw & AgentSkills harnesses | symlink/copy skills/mental-models/ into the harness's skills dir |
CLI-first, same content |
| CLI | pip install mental-models or uvx mental-models |
mental-models select "..." |
| MCP server | pip install mental-models-mcp or uvx mental-models-mcp |
add to Claude Desktop / Cursor / Zed / Continue / OpenCode / OpenClaw config |
| Python library | pip install mental-models |
from mental_models import select_models |
All five routes share the same latticework logic — the CLI is the single source of truth, and every other surface delegates to it.
Charlie Munger argued that worldly wisdom comes from building a latticework of mental models drawn from many disciplines—then hanging experience on that lattice. A single discipline gives you a hammer; a latticework gives you judgment. This skill operationalizes that idea: instead of one framework, Claude reaches across psychology, physics, economics, math, and strategy to analyze your problem. Read more in Poor Charlie's Almanack or Farnam Street's Mental Models hub.
CLI (fastest on-ramp — no install):
uvx mental-models select "how do I decide between two jobs" -k 5Or install it:
pip install mental-models
mental-models select "our team keeps missing deadlines" --jsonClaude Code skill — symlink (stays in sync) or copy:
git clone https://github.com/cyperx84/claude-skills-mental-models.git
ln -s "$(pwd)/claude-skills-mental-models/.claude/skills/mental-models" ~/.claude/skills/mental-modelsThen in Claude Code: "Apply inversion to my pricing strategy." The skill auto-activates on phrases like help me think, apply mental model, or any model name (inversion, bottlenecks, second-order thinking, margin of safety...).
Portable skill for OpenClaw & other AgentSkills harnesses:
# drop into your harness workspace
ln -s "$(pwd)/claude-skills-mental-models/skills/mental-models" <workspace>/skills/mental-modelsDetails and per-harness instructions: docs/openclaw/README.md.
MCP server (new in v0.2.0) — for Claude Desktop, Cursor, Zed, Continue, OpenCode, OpenClaw, or any MCP-capable harness:
pip install mental-models-mcp
# or: uvx mental-models-mcpThen drop the matching config snippet from docs/mcp/ into your client.
Python library:
from mental_models import select_models, get_model
for m in select_models("scaling issues", top_k=5):
print(m.slug, "-", m.name)Apply inversion to this architecture decision
What mental models help with scaling systems?
Help me think through whether to take this job offer
Full catalog in docs/categories.md.
| Category | Range | Sample |
|---|---|---|
| General Thinking | m01–m09 | First principles, inversion, second-order thinking |
| Science | m10–m29 | Leverage, inertia, feedback loops |
| Systems Thinking | m30–m40 | Bottlenecks, scale, margin of safety |
| Mathematics | m41–m47 | Randomness, regression to the mean |
| Economics | m48–m59 | Trade-offs, scarcity, creative destruction |
| Art | m60–m70 | Framing, audience, contrast |
| Strategy | m71–m75 | Asymmetric warfare, seeing the front |
| Human Nature | m76–m98 | Cognitive biases, incentives, social proof |
Mermaid source
%% Latticework core (top 15 hubs)
graph LR
m01["The Map Is Not the Territory"]:::c_General_Thinking_Too
m07["Inversion"]:::c_General_Thinking_Too
m14["Ecosystems"]:::c_Physics__Chemistry__
m15["Evolution (Natural Selection and Extinction)"]:::c_Physics__Chemistry__
m17["Friction and Viscosity"]:::c_Physics__Chemistry__
m20["Inertia"]:::c_Physics__Chemistry__
m21["Leverage"]:::c_Physics__Chemistry__
m31["Equilibrium"]:::c_Systems_Thinking
m34["Margin of Safety"]:::c_Systems_Thinking
m46["Surface Area"]:::c_Mathematics
m49["Supply and Demand"]:::c_Economics
m50["Optimization"]:::c_Economics
m52["Specialization"]:::c_Economics
m54["Efficiency"]:::c_Economics
m55["Debt"]:::c_Economics
m17 --- m20
m34 --- m55
m50 --- m54
m07 --- m34
m07 --- m50
m07 --- m54
m17 --- m50
m31 --- m46
m01 --- m07
m01 --- m31
m01 --- m34
m01 --- m55
m07 --- m17
m07 --- m31
m07 --- m46
m07 --- m55
m14 --- m31
m14 --- m49
m15 --- m49
m15 --- m52
m17 --- m21
m17 --- m54
m21 --- m31
m21 --- m34
m21 --- m55
m31 --- m34
m31 --- m55
m01 --- m21
m01 --- m49
m01 --- m50
m01 --- m52
m07 --- m21
m07 --- m52
m14 --- m15
m14 --- m20
m14 --- m46
m17 --- m46
m20 --- m21
m20 --- m46
m20 --- m50
m20 --- m52
m20 --- m54
m21 --- m50
m31 --- m52
m34 --- m46
m46 --- m52
m46 --- m54
m46 --- m55
m49 --- m55
m50 --- m55
m54 --- m55
classDef c_Economics fill:#1f2937,stroke:#fbbf24,color:#fff
classDef c_General_Thinking_Too fill:#1f2937,stroke:#60a5fa,color:#fff
classDef c_Mathematics fill:#1f2937,stroke:#a78bfa,color:#fff
classDef c_Physics__Chemistry__ fill:#1f2937,stroke:#34d399,color:#fff
classDef c_Systems_Thinking fill:#1f2937,stroke:#f472b6,color:#fff
Auto-generated from shared-keyword analysis. Full graph: docs/latticework.json.
One problem, many lenses—that's the point.
Two sibling packages ship the latticework to any harness — no Claude Code required:
packages/mental_models/— the CLI and Python library. This is the single source of truth:mental-models select | get | list | categories | apply | which | doctor | version, all with--json. The Claude Code skill and MCP server both shell out to it.packages/mental_models_mcp/— a Model Context Protocol server (new in v0.2.0) that exposes the same selector to Claude Desktop, Cursor, Zed, Continue, OpenCode, and any other MCP client.
pip install mental-models # CLI + library
pip install mental-models-mcp # MCP serverfrom mental_models import select_models, get_model
for m in select_models("how do I decide between two jobs", top_k=5):
print(m.slug, "-", m.name)
print(get_model("inversion").description)See packages/mental_models/README.md for the full API.
Model additions, fixes, and new examples welcome. See CONTRIBUTING.md.
Sources, citations, and further reading in REFERENCES.md.