What problem does it solve? Numerical Python code often fails silently: optimizers return unconverged vectors, root solvers miss brackets, and sparse matrices get accidentally densified. This Skill enforces explicit execution contracts, residual verification, and failure routing so SciPy results are provably correct rather than assumed correct. ## Core Features & Use Cases - Solver selection guidance: Choose optimization, root-finding, integration, interpolation, sparse, signal, spatial, or statistical routines based on smoothness, bounds, derivative availability, and sparsity. - Evaluated solution recipes: Anchored implementations for bounded minimization with gradient checks, bracketed scalar root solving, and sparse linear solves with residual guards, each paired with a verification test. - Verification and failure routing: Classify failures as input-contract, environment, execution, or invariant issues, and reject nonconvergence, nonfinite outputs, and excessive residuals before declaring completion. - Use Case: When debugging a scipy.optimize.minimize call that returns a result without converging, the Skill directs you to check the success flag, verify the gradient residual independently, and scale tolerances from the application's units. ## Quick Start Ask the assistant to solve a bounded least-squares problem with scipy.optimize.minimize and verify the optimum with an independent residual check.
