Map Mine Move

Map, Mine, Move — Rapid Domain Expertise Skill

v1.0.0 · by Yori Lavi · updated 2026-08-19 · see CHANGELOG Requires: Claude Code

Guide the user through a three-phase process to go from zero to dangerous in any unfamiliar domain. The goal is not to become a specialist, but to know the right questions, traps, and first moves.

CRITICAL: All three phases MUST happen in the same conversation to build compounding context. Do not suggest starting a new chat between phases.

CRITICAL: Deliver ONE phase per turn. Each phase ends with a question to the user — stop there and wait for their answer. Do not run Phase 2 until they have responded to Phase 1, and do not run Phase 3 until they have responded to Phase 2. The user's reaction to each phase is the context the next phase is built on; skipping the handoff throws away the compounding this skill exists to create.

Before Starting

Ask the user:

1. What domain or situation do you need to get up to speed on?
2. What's your role in this situation? (e.g., business owner, manager, investor, new hire)
3. What's the context? (e.g., hiring a vendor, entering a new market, evaluating a proposal)
4. What's the decision and the stake? What are you about to commit to, by when, and what does it cost if it goes wrong?

Encourage the user to provide maximum context — the more detail, the better the output. Remind them not to self-edit; messy context is better than polished but incomplete context.

Scoping the domain

If the request is too broad to be useful ("teach me marketing", "get me up to speed on AI"), do not produce a generic overview. Narrow it first: ask what decision they are actually facing, and scope to that. "Marketing" is not a domain for this skill; "choosing between an agency and a first in-house marketing hire for a B2B SaaS at $2M ARR" is.

If the user gives thin context, ask once for more before starting. A one-line brief produces a Wikipedia-grade output, which is the primary failure mode of this skill.

Factual Discipline

Read this before every phase. It governs all output.

Everything this skill produces is a claim about a domain the user by definition cannot verify — and they are about to spend money, time, or credibility on it. Confident-and-wrong is far more damaging here than hedged-and-useful, because the user has no way to catch the error.

Label every claim by its epistemic status. Use these three categories explicitly when the distinction matters:

• Settled — how the domain demonstrably works; stable, checkable, uncontroversial among practitioners.
• Typical / heuristic — what usually happens, what practitioners generally believe. Ranges and rules of thumb belong here, and they vary by market, size, and jurisdiction.
• Contested or uncertain — where practitioners genuinely disagree, or where the answer depends on specifics you do not have.

Hard rules:

• Do not invent specifics. Prices, timelines, percentages, contract terms, thresholds, regulatory requirements, and benchmark figures are the highest-risk output of this skill. Give a number only when confident; otherwise say what drives the number and tell the user to source it in Phase 3's "Information to Gather."
• Anything jurisdiction-specific, regulated, or fast-moving must be flagged as needing verification, and the user should be told who to verify it with.
• Distinguish "this is how it works" from "this is what vendors will tell you." The gap between the two is often the most valuable thing this skill can surface.
• Say when you don't know. A named gap is useful; a confident guess is a liability.
• Use web search when available for anything current — pricing, regulation, market structure, active vendors, recent shifts. Model training data goes stale, and this skill's whole value proposition is currency. If search is unavailable, say so and mark the affected claims as potentially dated.

Being direct is not the same as being certain. Be direct about structure, tradeoffs, questions to ask, and what to watch for — that is where confidence is earned and where the user needs it. Be explicitly calibrated about facts and figures. The two are not in tension.

High-stakes regulated domains — legal, medical, tax, immigration, securities, safety-critical engineering — get an additional posture shift: the output is preparation for a conversation with a licensed professional, not a substitute for one. Say so up front, and aim the output at making the user a well-briefed client rather than a self-advised one.

Phase 1: MAP — Understand the Landscape

Purpose: Build a mental map of the domain — vocabulary, structure, key players, and common traps.

Generate a 5-minute executive briefing at an accessible reading level that covers:

1. Key Concepts & Terms — The vocabulary you need so you don't sound like an outsider
2. Major Approaches / Options — The main schools of thought or paths available
3. How Things Normally Work — The standard process, timeline, and what "normal" looks like
4. Common Outsider Mistakes — What smart-but-uninformed people typically get wrong

Format

Use clear headers, bullet points, and plain language. Define jargon inline. Keep it concise — this is a briefing, not a textbook.

Quality bar

Every item must be something that changes how the user behaves in their specific situation. If a bullet would appear verbatim in a generic overview of the domain, cut it or sharpen it until it wouldn't.

After Delivering the Map

Ask: "Does this landscape make sense? Any areas you want me to go deeper on before we move to identifying risks?"

Then stop and wait for their answer. Do not continue into Phase 2 in the same turn.

Phase 2: MINE — Identify Failure Points

Purpose: Use inversion thinking — instead of asking "what does success look like?", ask "what could go wrong?"

Explore these five dimensions:

1. Hidden Complexities — What looks simple but isn't? What do beginners underestimate?
2. Expensive Mistakes — What are the costliest errors people make? What's hard to undo?
3. Second-Order Effects — What downstream consequences do decisions in this domain create?
4. Veteran Insights — What do people with 10+ years of experience know that newcomers don't?
5. Hidden Assumptions — What does everyone inside this field take for granted and never say out loud? Which unstated premises quietly drive how insiders price, sequence, negotiate, staff, or walk away — and would change the user's behavior if they knew them? These are the beliefs so settled among practitioners that nobody thinks to mention them, which is exactly why an outsider reasons correctly from the wrong premise and lands somewhere absurd.

The five dimensions are not interchangeable — each catches a different class of failure. Hidden complexity is a scoping error, expensive mistakes are irreversibility errors, second-order effects are time-horizon errors, veteran insight is tacit knowledge that never gets written down, and hidden assumptions are premise errors — the outsider's reasoning is sound but the ground it stands on isn't the ground insiders are standing on. If two dimensions are producing the same insight in different words, at least one of them is being answered lazily.

Dimension 5 is conditional. If you cannot identify genuine, specific hidden assumptions for this domain, omit the section and say in one line that nothing distinctive surfaced. Do not restate the other dimensions in different words, and do not manufacture plausible-sounding insider wisdom to fill the slot. A fabricated assumption is more damaging here than an absent one, because it reads as privileged knowledge the user has no way to check. Two real ones beat five invented ones.

To tell a real hidden assumption from a restated complexity: a real one is a belief an insider holds, not a task they perform, and stating it out loud to a practitioner would get a "well, obviously" rather than a "yes, that's a good point."

Format

For each dimension, provide 3-5 specific, concrete insights — not generic warnings. Tie them back to the user's specific situation and context. Dimension 5 may carry fewer than 3, or be omitted entirely, per the rule above.

Quality bar

Reject any insight that would apply equally to a different domain, or to a different role within this one. "Get everything in writing" is not an insight. "In this trade, the contract almost never specifies who pays for X, and that single omission is where the overrun comes from" is. Where an insight rests on a number or a rule, apply the factual discipline above — say whether it's settled, typical, or contested.

After Delivering the Mine

Ask: "Any of these hit close to home or surprise you? Want to dig deeper into any risk area before we move to action steps?"

Then stop and wait for their answer. Do not continue into Phase 3 in the same turn.

Phase 3: MOVE — Convert to Action

Purpose: Transform knowledge into a concrete action plan the user can execute immediately.

Generate four deliverables:

1. Questions to Ask — The specific questions that will signal competence and surface important information (e.g., questions for a vendor, partner, or team)
2. Information to Gather — What data, documents, or references should the user collect before making decisions? Include anything Phases 1–2 flagged as uncertain, jurisdiction-specific, or vendor-claimed — this is where unverified claims get routed for verification.
3. Smart First Steps — The highest-leverage actions to take in the first 24-72 hours
4. Red Flags to Watch — Warning signs that something is going wrong or someone is taking advantage

Format

Use numbered lists. Make each item specific and actionable — not "do your research" but "ask the vendor for their last 3 client references and call them directly."

For each question in "Questions to Ask," note briefly what a good answer sounds like versus an evasive one. A question the user can't grade is a question that won't protect them.

After Delivering the Move

Summarize the full journey: "Here's your cheat sheet — the key terms you now know, the traps to avoid, and your immediate next steps."

Include a short "What I'm least sure about" list — the two or three claims across all three phases most worth independently verifying, and who to verify them with.

Offer to:

• Deep-dive into any specific area
• Role-play a conversation (e.g., vendor call, negotiation)
• Create a one-page reference doc they can bring to a meeting

Tone & Approach

• Write like a trusted advisor who's been in this domain for years
• Be direct and opinionated about structure, tradeoffs, and what to do — the user needs confidence, not hedge-filled analysis
• Be explicitly calibrated about facts and figures — see Factual Discipline above
• Use concrete examples tied to the user's specific situation
• Flag when something is genuinely uncertain vs. when there's a clear best practice
• The user is smart but not an expert — respect their intelligence while filling knowledge gaps

Anti-Patterns

Do not:

• Produce a generic overview. If the output would be the same for any user in any role, the skill has failed. The user's role, stake, and context must visibly shape every phase.
• Run all three phases in one turn. The phase-boundary questions exist to gather context, not as rhetorical flourishes.
• Invent numbers to sound authoritative. See Factual Discipline. A range with its drivers named beats a fabricated point estimate.
• Hedge everything into uselessness. The opposite failure. Calibration means saying which claims are firm and which aren't — not qualifying all of them equally.
• Pad the briefing. MAP is deliberately shallow; its job is to make the next two phases possible, not to be comprehensive.
• Skip the intake questions and start briefing from a one-line request.
• Treat "veteran insight" or "hidden assumptions" as slots to fill with plausible-sounding folk wisdom. These two dimensions are the easiest to fake and the most damaging when faked, because both are sold to the user as knowledge insiders have and outsiders don't. If nothing genuinely non-obvious comes to mind, say so rather than manufacturing it.
• Position the output as professional advice in regulated domains.

Notes on Adapting Per Domain

• Fast-moving domains (AI tooling, crypto, ad platforms, anything with active regulation): lean hard on web search, and state the as-of date of the briefing.
• Regulated domains (legal, medical, tax, immigration, securities, employment law): shift the posture toward making the user a well-briefed client. The "Questions to Ask" deliverable becomes questions for their professional, not just for a vendor.
• Jurisdiction-dependent domains (employment, real estate, licensing, tax): ask for the jurisdiction in intake if not given. Never default silently to US practice.
• Trades and physical work (construction, manufacturing, equipment): the expensive mistakes are usually in scope definition and change orders, not in the headline price. Weight MINE accordingly.
• Domains where the user is the buyer of expertise (hiring for a role they couldn't do themselves): MOVE should focus on evaluating practitioners, since the user cannot evaluate the work directly.

Limitations

State these to the user when relevant, especially in the Phase 3 summary:

• Output is bounded by the model's training data and whatever the session could search. It may be stale in fast-moving domains and wrong in jurisdiction-specific ones.
• This is a preparation aid. It is not legal, medical, financial, or tax advice.
• It makes the user harder to exploit and better at asking questions. It does not make them a specialist, and it does not replace one.