thinking-model-router

# Thinking Model Router Use this skill before complex analysis so the system chooses the right reasoning frame instead of applying generic thinking. ## Core Rule Pick one to three models. Do not use all models at once unless the task is high-stakes and explicitly needs a multi-model review. ## Dispatch Table | Situation | Use | |---|---| | Need to rebuild from fundamentals | First principles | | Need to inspect downstream effects | Second-order thinking | | Need to avoid failure | Inversion or pre-mortem | | Need structured decision analysis | Kepner-Tregoe | | Need to know whether to move fast | Reversibility | | Need to prioritize scarce resources | Opportunity cost | | Need to update beliefs from weak evidence | Bayesian thinking | | Need to reduce biased judgment | Debiasing or steel-manning | | Need a good-enough decision under constraints | Bounded rationality | | Need to clarify vague requirements | Socratic questioning | | Need to reason under uncertainty | Probabilistic thinking | | Need to understand components and side effects | Systems thinking | | Need to find reinforcing or balancing dynamics | Feedback loops | | Need to detect recurring organizational patterns | System archetypes | | Need fast action under changing conditions | OODA | | Need high-impact intervention | Leverage points | | Need bottleneck optimization | Theory of Constraints | | Need to classify complexity | Cynefin | | Need simpler explanations | Occam's Razor | | Need to separate metrics from reality | Map-territory | | Need to decide what to own or escalate | Circle of competence | | Need to resolve a contradiction | TRIZ | | Need root-cause analysis | Five Whys Plus | | Need testable debugging | Scientific method | | Need edge-case exploration | Thought experiment | | Need rough sizing | Fermi estimation | | Need risk buffer | Margin of safety | | Need durable method selection | Lindy effect | | Need improvement by removal | Via negativa | | Need to attack the plan before launch | Red team | | Need product/customer clarity | Jobs to Be Done | | Need to start from available resources | Effectuation | | Need model choice itself | Model selection or model combination | ## Workflow 1. Classify the task: - decision, debugging, strategy, product, risk, system, estimation, or discovery. 2. Select one to three models from the dispatch table. 3. State why each selected model fits. 4. Apply the models in a useful order: - discovery models first, - decision models second, - risk models before execution, - verification models at the end. 5. Convert the output into an action, test, artifact, or decision record. ## Recommended Combinations - AI transformation: Cynefin, Systems Thinking, Theory of Constraints, Pre-mortem. - Debugging: Scientific Method, Five Whys Plus, Occam's Razor. - Product design: Jobs to Be Done, Map-territory, Red Team. - Architecture choice: Reversibility, Lindy Effect, Margin of Safety. - Marketing strategy: Opportunity Cost, Second-order Thinking, Bayesian Thinking. - Workflow automation: Theory of Constraints, Via Negativa, Effectuation. ## Falsifiability Checks Every model must produce a checkable claim: - What observation would prove this wrong? - What is the cheapest test? - What evidence would change the decision? - What output will show that the model helped? ## Anti-Patterns - Listing many models without applying them. - Using a model because it sounds impressive. - Skipping evidence after a model produces a conclusion. - Treating dashboard metrics, reports, or AI output as reality without checking ground truth. ## Portability Contract This skill must be usable from any connected IDE, CLI, MCP client, or agent runtime through SEOSONA OS portable routing. - Discover through `2_KNOWLEDGE/SKILLS_ROUTER.md` or `1_CORE/scripts/seosona_capability_bridge.js`. - Reference system files with `~/.seosona`, `${SEOSONA_ROOT}`, or relative paths. - Do not depend on the physical installation path or the environment that originally ingested the source material. TASK COMPLETED