A one-file skill for Claude Cowork that makes Claude reframe your question, challenge your assumptions unprompted, think several moves ahead, and land one calibrated recommendation even when information is incomplete. Modeled on The Professor from Money Heist: the cognitive discipline, not the character. Installs in 60 seconds; scroll down 10 lines to see exactly what it changes.
Pairs with the Decision Frameworks skill for high-stakes calls.
Same question. Same Claude. Two different answers.
Question: "Should I quit my job and go full-time on my side project?"
That's an exciting decision! There are several factors to consider. First, assess your financial runway. Second, evaluate the market opportunity. Third, consider your personal risk tolerance. Fourth, think about your support system...
Ultimately, this is a personal decision that depends on your unique circumstances. I'd recommend making a pros/cons list and consulting with a financial advisor.
Structured. Safe. Useless. A framework when you needed a read.
Question: "Should I quit my job and go full-time on my side project?"
You're framing this as a binary. It's actually a sequencing problem. The real question is: what's the minimum viable proof this works before you burn the safety net?
The assumption I'd challenge: You're treating quitting as the accelerant. In most cases, external pressure doesn't create clarity; it creates panic. Execution doesn't improve under survival pressure. It gets defensive.
What you haven't asked: Is this a revenue problem or a time problem? Because those need completely different interventions. If you're at $2–3K/month on the side, the constraint is probably time, and quitting might be right. If you're not there yet, quitting fixes the wrong variable.
The move: Don't quit. Set a target: 3 months, $X revenue, a specific milestone. Hit it, and you quit with data. Miss it, and you learn what 40 extra hours/week actually buys before betting your income on it.
What's current monthly revenue, and what does "full-time" let you do that you can't do now?
That reply reframed the question, named the driving assumption, surfaced the thing you hadn't asked, and gave one concrete recommendation instead of a framework.
See DEMO.md for 3 full before/after examples across career, business strategy, and plan evaluation.
| If you are... | Mastermind helps you... |
|---|---|
| Founder / builder | Cut through your own confirmation bias before committing to a direction |
| Consultant / freelancer | Give clients the uncomfortable truth instead of the comfortable answer |
| Career switcher | Make irreversible moves (quit, pivot, relocate) with evidence instead of anxiety |
| Content creator | Evaluate your strategy rather than just execute it |
| Anyone with a hard decision | Get one calibrated recommendation instead of a list of things to "consider" |
Not useful for factual lookups, syntax questions, casual conversation, or emotional support with no decision attached.
The skill encodes a thinking protocol that runs silently before every response:
- Reframe before solving. The stated problem is usually the symptom.
- Challenge assumptions unprompted. Especially the ones the user believes most.
- Think downstream. Map what happens next if the user acts on the best answer.
- Surface the unasked. The risk they haven't named, the dependency they haven't clocked.
- Read emotional context as strategic data. Validation-seeking disguised as a question gets named.
- Stay direct, not harsh. Name the flaw without softening a correct assessment.
- Ground load-bearing facts. Tags confidence: Verified / Reasoned / Assumed / Unknown.
- Always give a calibrated recommendation. Never ends on "it depends."
- Apply the high-stakes guardrail. Routes to experts for medical/legal/financial calls.
- No framework inflation. Depth matches complexity; simple answer for a simple problem.
| Behavior | Standard Claude | Mastermind | Mastermind + Decision Frameworks |
|---|---|---|---|
| Opens with | Restates your question | Reframes it | Reframes + gates the facts first |
| Wrong premises | Answers around them | Names them directly | Names them + runs /premortem to find what else is wrong |
| Options given | Lists 4–6 equally | One recommendation | One recommendation + /scenarios for the branches |
| Uncertainty | "It depends" + caveats | Calibrated call + what changes it | Calibrated call + base rates + adversarial stress test |
| Unasked issues | Rarely surfaces them | By design, every time | By design + /blindspot for structured unknown-unknown scan |
| Best for | General questions | Tactical to mid-stakes decisions | Irreversible, high-stakes, or adversarial situations |
See DECISION_FRAMEWORKS.md for the combined protocol and a full worked example.
Your question
│
▼
/mastermind
(Reframe → Challenge assumption → Surface unasked → Recommend)
│
├─── Moderate stakes, reversible ──→ Done. Act on it.
│
└─── High stakes / irreversible ──→ Escalate
│
▼
/gate (Are the facts solid enough?)
│
▼
/baserates (What usually happens here?)
│
▼
Pick 1–2 frameworks:
/premortem → What kills this?
/redteam → How does an adversary break it?
/blindspot → What haven't we considered?
/scenarios → What if the future looks different?
│
▼
Final calibrated call with confidence tags
Requires the Claude Cowork desktop app.
Install Mastermind:
- Download
mastermind.skillfrom this repo - Open Claude Cowork → Settings → Capabilities → Skills
- Click Install from file → select
mastermind.skill - Done. Type
/mastermindin any session.
Install Decision Frameworks (optional, recommended for high-stakes decisions):
- Download
decision-frameworks.skillfrom this repo - Same path: Settings → Capabilities → Skills → Install from file
- Open Claude Cowork → Settings → Capabilities → Skills → Create skill
- Name it
mastermind - Copy the contents of
SKILL.mdinto the skill prompt - Save
/mastermind
think like the professor
what am I missing
stress-test this
what's my play
challenge my thinking
help me think through
what's the real problem here
For high-stakes, irreversible decisions, pair this skill with the Decision Frameworks skill: 20 structured analytical tools that run when Mastermind's fast read isn't enough.
Mastermind handles the fast strategic read (reframe, challenge, recommend). Decision Frameworks handles the deep audit (premortem, red team, base rates, war-gaming, blindspot scan).
Install both, then use this sequence:
1. /mastermind → Fast read. Most decisions end here.
2. /gate → Are the facts good enough to analyze?
3. /baserates → What usually happens to cases like this?
4. /premortem → What kills this if we commit?
5. /redteam → How would an adversary break it?
Full protocol, 20 commands, and a worked example in DECISION_FRAMEWORKS.md.
| Use Mastermind for | Skip it for |
|---|---|
| Career moves (quit / stay / pivot) | Factual lookups |
| Business planning (pricing, launch sequence) | How-to syntax questions |
| Evaluating plans (finding the weak assumption) | Simple one-step tasks |
| Cutting through over-complication | Emotional support with no decision |
| Interpersonal situations | Casual conversation |
Modeled on Sergio Marquina (The Professor) from Money Heist. The thinking method, not the heist:
- He answered the real question under the surface question
- He read people and situations with equal precision; emotional context was strategic data
- He had Plan A, B, and C before others finished Plan A
- He challenged every assumption in his own plan, especially the ones he believed most
- He never let urgency collapse thinking into reaction
One correction built into the skill: The Professor's confidence came from months of verified preparation. This skill replicates the method, not the cinematic certainty. Confidence is a function of evidence, not delivery.
| File | What it is |
|---|---|
README.md |
This file: overview, before/after, install guide |
SKILL.md |
Full skill prompt, copy this for manual install |
mastermind.skill |
Installable .skill file for Claude Cowork |
decision-frameworks.skill |
Decision Frameworks skill, install alongside Mastermind for the full stack |
DEMO.md |
3 full before/after examples (career, business, plan evaluation) |
DECISION_FRAMEWORKS.md |
Combined Mastermind + Decision Frameworks protocol + worked example |
LICENSE |
MIT |
MIT. Use it, adapt it, build on it.
Saurabh Sarkar, AI Workflow Automation github.com/Sroyal7 · saurabhsarkar.7397@gmail.com