Load any thinker's mental operating system into your AI. Not role-play. Not impersonation. Cognitive architecture transfer.
You've tried "act like Elon Musk" or "think like a Stoic philosopher."
The AI gives you a costume — some quotes, a vague tone, maybe a catchphrase. It doesn't give you the operating system. It doesn't give you how they actually frame problems, what they optimize for, what psychological wounds drive their decisions, or where their thinking reliably breaks down.
That's the gap this skill fills.
Thinking Framework performs a deep, multi-layer excavation of any target:
Cognitive Layer (C1–C7)
- Mental models they use to explain reality
- Decision heuristics that drive choices under pressure
- What they're actually optimizing for (stated vs. hidden)
- How they reframe problems before solving them
- Risk posture and time horizon architecture
- Paradox and contradiction structure
Psychological Layer (P1–P8)
- Core wound and formative architecture
- Dominant motivational deep structure
- Ego structure and identity architecture
- Defense mechanisms under pressure
- Shadow profile (what's suppressed / projected)
- Relational and attachment patterns
- Existential relationship (mortality, legacy, meaning)
- Stress and breakdown profile
Once loaded, the AI reasons FROM inside that framework — not about it.
| ❌ Standard role-play | ✅ Thinking Framework |
|---|---|
| "Elon Musk would say: move fast and break things!" | Strips the analogy. Finds the cost floor. Attacks the bottleneck. |
| "A Stoic would advise you to be calm." | Immediately partitions: what's in your control? Releases the rest without discussion. |
| "Warren Buffett believes in long-term investing." | "Will this business exist in 20 years? If the answer is uncertain, nothing else matters yet." |
User: "I work hard but can't reach the success I want. Should I quit or keep going?"
Marcus Aurelius framework loaded:
The question itself is wrong. "Quit or continue" assumes the framing is correct. First partition: what's in your control? The quality of today's work. The clarity of your direction. Not the timing of results. Not external recognition. You're measuring yourself against something you don't control — that's what's exhausting you, not the work itself.
Elon Musk framework loaded:
Stop. What exactly are you building? "Success" isn't measurable. Where is the bottleneck in your system? You're probably working hard downstream of a blockage that makes all that effort irrelevant. Find the constraint first. Everything else is noise until you do.
Same question. Two entirely different cognitive architectures. Both genuinely useful.
- Open OpenClaw
- Go to Skills → Browse → search "thinking-framework"
- Install → done
- Download or clone this repo
- In OpenClaw: Settings → Skills → Import folder
- Select the
Thinking-Frameworkfolder
Simply tell your AI:
"Load the thinking framework of [target]"
"Think like [target]"
"Activate [target] mindset"
"[target] mode"
"How would [target] approach [problem]?"
"Blend [target A] with [target B]"
Works with any person, philosopher, organization, discipline, or school of thought.
After loading, the AI presents a Framework Card summarizing the full cognitive and psychological architecture, then stays in that mode until you exit.
Exit anytime:
"Exit framework"
"Switch to [new target]"
"Blend [target] with current framework"
"What does this framework miss here?"
For a full list of prompt patterns and tips, see WORKFLOW.md.
Anything with enough documented evidence:
- Individual thinkers & leaders — philosophers, founders, investors, generals, scientists
- Organizations — companies, military units, research institutions
- Philosophical schools — Stoicism, Zen, Existentialism, Austrian Economics
- Canonical texts & systems — specific books, methodologies, frameworks
- Disciplines — physics thinking, legal reasoning, game theory, design thinking
Thinking-Framework/
├── SKILL.md ← Skill specification & execution protocol
├── WORKFLOW.md ← User guide: prompt patterns & tips
├── CHANGELOG.md ← Version history
├── references/
│ ├── layer1-cognitive.md ← 7-dimension cognitive excavation engine
│ ├── layer2-psychological.md ← 8-dimension psychological deep analysis
│ ├── layer3-operational.md ← Active framework mode behavioral protocol
│ └── model-guidance.md ← Compensation for weaker / local models
└── examples/
├── richard-feynman.md ← Full worked example: complete excavation
└── composite-feynman-aurelius.md ← Composite mode: Feynman × Aurelius
| Model | Quality |
|---|---|
| Claude Opus / Sonnet | ⭐⭐⭐⭐⭐ Full depth — all 15 dimensions |
| GPT-4o+ | ⭐⭐⭐⭐⭐ Full depth |
| Gemini Pro | ⭐⭐⭐⭐ Strong — psychological layer slightly lighter |
| Llama 70B+ | ⭐⭐⭐ Good — cognitive layer strong, psychology moderate |
| Local 7B–13B | ⭐⭐ Basic — cognitive layer only |
The skill includes references/model-guidance.md with explicit compensation
instructions for weaker models — it adapts automatically.
This skill is designed for cognitive analysis, not impersonation.
- Never generates content formatted as real quotes from real people
- Never fabricates biographical facts or undocumented decisions
- Always labels inferences separately from documented facts
- Surfaces blind spots and failure modes — honest analysis, not flattery
- Living persons: stays within publicly documented cognitive and behavioral patterns only
Issues and PRs welcome. Particularly interested in:
- Additional worked examples (new targets, composite blends)
- Additional layer reference files (e.g.,
layer4-sociological.md) - Organization and discipline-specific analysis templates
- Test cases and example outputs for different models
MIT-0 — Free to use, modify, and redistribute. No attribution required.