thinking-framework-skills

Builds an abstraction ladder that moves a problem up ("why / to what end?") and down ("how / what specifically?") to locate the right altitude to work at, then marks one rung as the working level. Use when a problem is stated as a bare solution with an unstated purpose, as a vague aspiration with no concrete handle, when people are arguing past each other at different levels, or before committing effort at an altitude nobody chose on purpose.

Produces a clustered theme map that groups many raw notes, observations, quotes, or data points bottom-up into a small set of named, traceable themes (the KJ method). Use when a scattered pile of dozens to hundreds of existing items needs to become a few emergent themes, such as synthesizing user-research notes, support tickets, survey free-text, or retro stickies, and the right structure should emerge from the data rather than be imposed.

Produces a structured after-action review by comparing what was expected against what actually happened, diagnosing why the gaps occurred, and converting them into specific owned sustain-and-change actions. Use when a project, launch, sprint, or incident has finished and you want to turn the outcome into learning, not a status update.

Handles a by-name request for ACH (Analysis of Competing Hypotheses), the evidence-by-hypothesis disconfirmation matrix, honestly. Controlled trials found ACH raises confidence with no accuracy gain and does not reduce confirmation bias, so this skill does NOT build the matrix as if valid. It leads with that evidence, then routes to the evidence-based move the job actually needs (think-red-team-light, think-evidence-vs-inference-sort, or think-what-would-have-to-be-true). Use only when someone asks for ACH or a competing-hypotheses matrix by name.

Produces an argument map by laying out a claim's supporting reasons, the co-premises each silently depends on, and the objections against it as an explicit structure, then flags the weakest links and unsupported premises. Use when an argument or recommendation must be evaluated for soundness, or when a fluent case may be hiding a broken inference.

Generates non-obvious ideas by surfacing the foundational assumptions a problem or solution rests on, negating each, and reframing from the reversed assumptions, then shortlisting, producing an assumptions-and-reversals sheet. Use when an option space feels stuck inside default constraints, or when you need to expose premises that are taken for granted.

Checks whether a decision has genuine minority dissent or only smooth surface consensus, identifies who actually holds a contrary view, and plans how to elicit and protect real dissent, flagging clearly where a view is constructed rather than authentically held. Use when consensus feels too easy, or to set up genuine challenge before a high-stakes call.

Produces a backcast path by fixing a vivid desired future state and reasoning backward through the milestones and preconditions required to reach it, ending at the next concrete step available now. Use when planning toward a transformative or long-horizon goal that forward planning anchors too low, when a chosen future needs a route mapped back to today, or when milestones and dependencies between now and the goal must be made explicit.

Produces a belief-update ledger that re-scores a standing inventory of open beliefs against newly arrived evidence on a cadence - each belief carrying a prior confidence, the evidence accrued since the last review, a revised confidence with an explicit delta and direction, a reason for the size of the move (a guard against under-updating), and a next-review trigger. Use when you hold consequential open beliefs, forecasts, or standing assumptions that should track new evidence over time and you want a disciplined, recurring re-score - not a one-time decision record, a finished-event retro, or a single claim's conditions.

Audits the boundary judgments that silently define a problem frame. For each of four sources, who benefits, who decides, whose knowledge counts, who has standing, it contrasts the descriptive is boundary against the normative ought boundary, names the gaps, and lists the affected-but-excluded parties who have no voice in the frame, producing a boundary-judgment audit rather than a stakeholder round-up. Use when a plan, metric, or proposal already encodes who matters and who does not and the risk is solving a tidy problem for the people inside the line while pushing harm onto people outside it, when an improvement claim rests on an unexamined judgment about whose improvement, or on contested, value-laden, multi-party situations.

Generates ideas the way silent parallel brainwriting does, producing several independent idea streams that build on each other without anchoring on the first voice, then consolidates them into a shortlisted idea pool. Use when you need breadth of options and want to avoid the anchoring and conformity that make ordinary brainstorming underperform.

Produces a four-layer matrix (litany, system, worldview, myth) of the current 'used future' beside a reconstructed preferred future per layer, anchored by a deliberately changed deep metaphor. Use when an issue is stuck because the framing is stuck, the official account and the system explanation have been argued to exhaustion, and the real disagreement is about clashing worldviews and the unexamined story underneath. Not a single-cause diagnosis (that is think-iceberg-model) and not a forecast.

Builds a signed causal loop diagram by closing the feedback loops in a situation, labeling each loop reinforcing (R) or balancing (B) with its link polarities, and reading likely dynamics (spiral, goal-seeking, or oscillation) off which loop dominates. Use when a situation feeds back on itself - growth that funds more growth, a fix that recreates its own problem, capacity that relieves then re-attracts demand - and you need to see why it keeps accelerating, stalling, or overshooting. Not for a single accumulation, a one-directional consequence tree, or a genuinely linear chain.

Runs a Complexity Domain Sort, sorting a situation into Clear, Complicated, Complex, Chaotic, or Confusion and choosing the response posture that fits, caveat-first. It leads with the weak evidence (the 2021 PMC internal-medicine review found scientific proof of its validity has yet to be provided), then forces the discipline the bare label lacks, namely an explicit response posture and a concrete next action rather than just a name. Use only when this complexity sort (also known as Cynefin) is asked for by name; for cause decomposition prefer think-issue-tree.

Builds a concept map - a non-hierarchical network of concept nodes joined by directed, labeled linking phrases so each node-link-node reads as an explicit proposition, with cross-links across clusters - then surfaces gaps, missing links, and questionable propositions. Use when a domain has many interrelated concepts and the goal is to externalize and inspect how they relate, forcing every relationship to be named rather than left as a vague association.

Produces a known-unknowns ledger by enumerating the relevant evidence not in hand, the variables that bear on a judgment but are unknown or unobservable, rating each unknown by its bearing and its obtainability, then re-rating confidence against the mapped gap. Use when a consequential one-off judgment rests on thin or partial evidence, no usable base rate exists, and the confidence on the table may be inflated by what was never looked at. Not for cases with a real reference class, for widening a numeric interval, or for unknowns cheap enough to just go resolve.

Converts a stuck trade-off into a contradiction to dissolve rather than a compromise to settle for. Names the two conflicting requirements, states an implementation-free Ideal Final Result, and applies the four separation principles (separate the requirements in time, space, scale, or condition) to satisfy both at once, then produces either a concrete resolution or an honest flag that the trade-off is genuine. Use when a decision has hardened into an either/or like speed versus quality, thorough versus fast, or generous versus profitable, when splitting the difference feels wrong and a win-win would be worth far more than the best compromise, or when a recurring design or product trade-off keeps forcing the same painful balance.

Produces a both/and polarity map - two interdependent poles, the upside and downside quadrant of each, a greater purpose and a deeper fear, early-warning signs, and action steps - for a chronic tension that should be managed rather than resolved. Use when an organization keeps oscillating destructively between two opposites that are both true and interdependent (centralize versus decentralize, stability versus change, candor versus diplomacy), and treating each swing as a problem to solve is the actual mistake. Not for a dissolvable trade-off and not for a real either/or choice.

Produces a decision journal entry that records a consequential decision at the moment it is made - the decision, the rationale, the predicted outcome, an explicit confidence level, and the assumptions it rests on - so it can be honestly reviewed later against what actually happened. Use when committing to a consequential, uncertain decision (a launch, hire, investment, bet, or strategic choice) and you want to lock in the prediction now to defeat hindsight bias and build calibration, or when pairing a record-now step with a later review.

Produces a criteria-weighted option matrix by comparing a set of options against weighted criteria, scoring each, surfacing the explicit tradeoffs, and recommending one while flagging where the scoring is soft. Use when choosing among several real options, or when a decision needs its tradeoffs made explicit rather than left to intuition.

Produces a dialectical estimate that improves a single committed number by polling the inner crowd, make a first estimate, assume it is wrong and list why, read whether that pushes the answer too high or too low, make a second estimate from those changed assumptions, then mechanically average the two. Use when a hard, one-off, bounded-scale numeric estimate (a year, a percentage, a count, a forecast) is about to be committed and no second judge, no real reference class, and no better data are available. Not for easy questions, not for unbounded order-of-magnitude unknowns, and never a cherry-picked single number in place of the average.

Runs one of the three lightweight prioritization presets (Eisenhower urgent-important 2x2, MoSCoW Must/Should/Could/Wont bucketing, or Pareto vital-few cut) caveat-first. It leads with the fact that this is a bundle of three different operations stapled by the word "prioritization", not one move, and each is weakly evidenced, then produces the one preset the user named with the discipline a canned template lacks. Use only when an Eisenhower matrix, MoSCoW list, or Pareto chart is asked for by name; for the rigorous versions prefer think-theory-of-constraints, think-one-way-vs-two-way-door, or think-decision-option-review.

Produces an ethical matrix, a grid that cross-references affected parties (rows, deliberately including voiceless ones such as future generations and the environment) against prima facie principles (columns of wellbeing, autonomy, fairness), specifying the option's impact in every cell and reading the result as a trade-off pattern rather than a score. Use when a concrete proposal affects multiple parties with a genuine moral trade-off among them, some affected parties have no voice, and an ethics debate keeps sliding between groups and principles. Not a decision calculator and not an option-versus-criteria scorer; the matrix maps the terrain and leaves the weighing to deliberation.

Produces an evidence/inference ledger by sorting the claims in a prompt, document, or proposed conclusion into evidence, inference, and assumption, attaching a confidence level to each inference and flagging anything uncited. Use when a recommendation must be trusted, or when you need to audit the reasoning behind a conclusion in a high-stakes context.

Evaluates competing actions under uncertainty by building a decision tree of choice and chance nodes, placing explicit probabilities on outcomes the decider does not control, rolling the tree back to an expected value per option, recommending the best-EV branch, and adding a what-flips-it note naming the probability or value that would reverse the choice. Use when a decision turns on uncertain outcomes you can put rough probabilities on, when the structure is sequential (a choice opens chance events that open later choices), and when the stakes justify making the probability assumptions explicit and inspectable instead of buried in a gut feel.

Generates novel solution candidates by stating a problem's deep relational structure, mapping it to distant source domains (nature, other industries, games), and transferring the mechanism rather than surface features, then adapting. Use when near, obvious solutions are exhausted and you need genuinely original approaches.

Produces a Fermi decomposition worksheet that estimates an unknown numeric quantity by factoring it into a chain of order-of-magnitude sub-estimates, guessing each to within a band, then multiplying back to a point estimate plus a compounded low/high range, with an independence check and a dominant-uncertainty flag. Use when you need a number and no lookup-able data or genuine reference class exists, so the magnitude has to be built from factors (for example sizing a market, a load, a cost, or a conversion you cannot look up). Not for forecasting from real base rates (use reference-class-forecasting) or decomposing a question for coverage with no number (use issue-tree).

Runs a Five Whys root-cause trace caveat-first. It leads with the weak evidence (Card 2017 found the single-chain method oversimplifies multi-causal problems), then forces the discipline a bare why-chain lacks, namely flagging at each step whether more than one cause could apply. Use only when Five Whys is asked for by name; for any problem that might have more than one cause prefer think-issue-tree.

Generates a new problem frame from theme analysis of the broader context, an abductive standpoint stated as approach it as if it were Y that redefines what the problem is and reasons forward to native solution directions. Use when an open paradoxical problem keeps resisting solutions inside the frame it arrived in, when conventional problem solving has only produced more of the symptom, or when the way a problem is framed is itself the obstacle and a fresh standpoint would be worth more than another solution attempt.

Produces a consequence map by tracing the first, second, and third order effects of a change or decision radiating outward from the center, surfacing ripples beyond the obvious and flagging the high-impact branches. Use when a decision has knock-on effects over time, or when first-order thinking is missing downstream risks and opportunities.

Produces an iceberg that moves a problem down four levels of causation - from the visible event, to the pattern over time, to the underlying structures, to the mental models that hold them in place - pairing each level with the intervention it implies to find systemic causes and higher-leverage fixes. Use when a problem keeps recurring despite event-level fixes, when a symptom is being treated as a one-off, or when the question is why this keeps happening and where to actually intervene.

Produces a negotiation preparation map that separates both sides' positions from their underlying interests, anchors the accept-or-walk decision on a named, valued best alternative inside the mapped zone of possible agreement, and invents options for mutual gain across differently-valued issues before dividing value against objective criteria. Use when agreement is required from a party not under your control and more than one issue is in play (salary, vendor, partnership, resource, or multi-party disputes), and a position fight has hardened before anyone checked whether the interests actually conflict. Preparation deskwork only, not live-table tactics, and not for a single-issue distributive haggle or a choice with no counterparty.

Audits whether a stated confidence interval means what it claims by running an equivalent-bet indifference test on its width and scoring nominal confidence against the actual hit rate, emitting corrected intervals and a calibration scorecard. Use when a human-stated interval or confidence number drives a consequential plan, forecast, or commitment and has never been audited, and the worry is overprecision rather than a wrong central estimate. Calibrates human-stated uncertainty only, never the agent's own confidence, and resizes width without relocating the number. Not for lookupable facts and not a promise of full debiasing.

Produces an issue tree that decomposes one big, ambiguous question top-down into a mutually-exclusive, collectively-exhaustive (MECE) set of smaller sub-questions, branch by branch, until the leaves are small enough to answer with data or judgment. Use when a question is too broad or multi-cause to answer as posed (for example "why is churn rising?", "where is margin leaking?", "should we launch a free tier?"), when analysis work must be split into non-overlapping parts, or when coverage matters and missing a whole category would be costly. Not for evaluating a given argument's soundness (use argument-mapping) or clustering existing notes (use affinity-mapping).

Produces an annotated reasoning trace that reconstructs how a conclusion was reached, from the observable data, to the data actually selected, to the meaning and assumptions added, then flags the riskiest leap and tests an alternative interpretation. Use when a conclusion feels certain but rests on interpretation, or to audit a contested inference.

Builds a simple mechanical scoring model - a few weighted predictive cues combined by a fixed formula and applied consistently - for a repeated predictive judgment, because consistent simple rules reliably match or beat holistic expert intuition. Use when the same kind of evaluative judgment recurs (screening candidates, scoring leads, triaging) and gut calls are inconsistent or overconfident.

Produces a regret (opportunity-loss) matrix with the minimax pick and the binding worst-case state, choosing among options over un-probabilized states of nature. Use when a few discrete options face a few discrete uncontrollable states and no defensible probabilities exist. Not an expected-value tree (which needs probabilities) and not an attribute scorecard.

Produces a morphological field (a Zwicky box of independent parameters by their possible values) plus a cross-consistency-pruned set of internally consistent configurations. Use when a solution is genuinely a configuration - a choice on each of several semi-independent dimensions (product architecture, service bundle, policy package) - and the risk is defaulting to one familiar combination and never seeing the rest of the space. Not a top-down MECE tree and not a provoked option list.

Converts a conditional-probability or base-rate question into natural frequencies over a concrete population (for example 9 of 1000) to compute the correct posterior and expose base-rate neglect, and refuses to proceed without real input rates. Use when interpreting a test result, screening signal, or any "given a positive, what is the real probability" question.

Produces a reversibility classification that triages a decision before any analysis - labeling it a reversible two-way door or a hard-to-reverse one-way door - and matches the deliberation and sign-off level to that verdict, so reversible calls are made fast and irreversible ones get real rigor. Use when it is unclear how much process a decision deserves, when a team is about to rubber-stamp something irreversible or convene a committee over something trivially reversible, or when a chronically slow org needs a defensible reason to move fast or slow down.