foundations-mathematical-optimization
Mathematical Optimization Foundations
Turn an allocation question into an explicit model and a defensible solution claim. Separate a good candidate, a feasible candidate, and a certified optimum.
When to Use
Trigger: constrained allocation, linear programming, convex optimization, integer programming, scheduling and assignment under capacity, solver choice, infeasible or unbounded models, duality certificates, optimality gaps, multi-objective tradeoffs, or robust/stochastic optimization.
Examples: allocate a fixed capacity across products; schedule GPU jobs under a GPU cap; find which constraints make a MILP infeasible; check an LLM-drafted model; verify an LP primal/dual witness.
Use decision theory to choose preferences or utilities, planning/search for action sequences with preconditions and effects (this skill owns resource-constrained scheduling), and theory of constraints to identify where improvement should focus. This skill owns the mathematical allocation after those choices. Ordinary prioritization without a quantitative constrained model need not activate it.