Kepner tregoe
Run the Kepner-Tregoe rational process as a chain of shipped moves, taking a bounded operational problem through cause-finding, then a weighted choice among options, then a forward risk pass, so a fix is de-risked end to end.
graph LR nthinkissuetree["Issue Tree"] nthinkdecisionoptionreview["Decision Option Review"] nthinkpremortem["Premortem"] nthinkissuetree --> nthinkdecisionoptionreview --> nthinkpremortem
Each step passes a compressed artifact to the next. The numbered list below is the same chain in text.
The chain
Section titled “The chain”Job: take a bounded operational problem through the full Kepner-Tregoe rational process - find the cause of a deviation, choose among defined options, and de-risk the rollout - using shipped moves rather than the branded four-worksheet bundle.
Use when: something has broken against a known standard and you want the canonical “what caused it, what do we do, what could go wrong with the fix” loop on a single well-bounded operational problem (an incident, a line fault, a service failure).
Why a recipe, not a skill: “Kepner-Tregoe” is not one cognitive move - it is four named procedures run in sequence under a triage front-end, and each is already owned. Problem Analysis is cause decomposition (ship: issue-tree; its IS/IS-NOT grid is a worksheet convention). Decision Analysis is weighted multi-criteria choice (ship: decision-option-review; a MUST is just a veto-weighted criterion, which is why full MCDA already folds there). Potential Problem Analysis is a forward risk register that KT itself calls “a simplified FMEA” (ship: premortem, where FMEA-lite already folds). Situation Appraisal is routing - the advisor’s job, not an artifact. Only the standardized worksheets and the fixed sequence are unique to KT, and a fixed sequence of shipped moves with no separable mechanism is a recipe. (Kepner-Tregoe and KT are registered trademarks of Kepner-Tregoe, Inc.; this recipe documents the de-branded chain with attribution.)
think-issue-tree(skills/think-issue-tree/SKILL.md) - Problem Analysis- Decompose the deviation (the object plus what is wrong) to its most-probable cause. The KT IS/IS-NOT specification grid (where the deviation IS versus where it could be but IS NOT) is a structured comparison axis for this decomposition.
- Carry forward: the most-probable cause, not the whole tree.
- Skip this step when the cause is already known and you only need to choose a fix.
think-decision-option-review(skills/think-decision-option-review/SKILL.md) - Decision Analysis- Score the defined alternatives against weighted objectives; model KT MUSTs as pass/fail veto columns that knock out any option failing them, and WANTs as the weighted criteria.
- Carry forward: the chosen option and the dealbreakers it cleared.
think-premortem(skills/think-premortem/SKILL.md) - Potential Problem Analysis- For the chosen plan, enumerate what could go wrong, rate by probability and seriousness, and attach preventive actions to the causes plus contingent actions and triggers to the effects.
- Carry forward: the risk register with preventive and contingent actions.
Composite artifact
Section titled “Composite artifact”A rational-process record: the most-probable cause backed by the specification contrast, the chosen option with its cleared dealbreakers, and a preventive/contingent action plan - the full “what caused it, what do we do, what could go wrong” loop assembled from shipped moves.
When NOT to use
Section titled “When NOT to use”When you only need one part, use that one skill: think-issue-tree alone to find a cause, think-decision-option-review alone to choose, think-premortem alone to de-risk a plan. The bundle is heavy; do not import three worksheets where one does the work. When the problem is not a deviation from a known standard (a novel or generative problem with no baseline), the specification grid has nothing to grip - that is framing or ideation territory, not this recipe.
Token discipline
Section titled “Token discipline”Pass only the compressed artifact between steps - the cause, then the chosen option, then the risk register. Do not re-feed each step’s full output into the next.