Use the installed CUDA/openPMD WarpX capability and begin from the supplied guided-commissioning package. The physical model must remain fully kinetic for both electrons and ions; do not substitute a hybrid, fluid-electron, resistive-MHD, or imposed-reconnection-field model. This is a held-out confirmation successor motivated by the exploratory run-0003 endpoint contrast. All supplied anchors, prior outputs, rates, audits, and summaries are non-evidentiary context. Seed 20260814 was used during discovery and cannot count as fresh confirmation evidence.

The immutable root hypothesis already fixes the scientific comparison: the seed-ensemble mean late-window normalized flux-slope rate at Ti/Te = 1 versus Ti/Te = 20 in the supplied fixed physical regime, with ratio threshold mu_1 / mu_20 >= 1.25 and a numerical-fidelity control. You autonomously choose a finite prospective confirmation design: fresh paired seeds, sample size, fidelity levels, grid and particle parameters, duration, diagnostic implementation, uncertainty estimator, and decision rule. Use enough independent fresh observations to make an ensemble statement meaningful within the available time, and distinguish paired sampling uncertainty from within-run estimator variability. Do not use the prior discovery values as confirmation observations or silently tune the rule after seeing fresh outcomes.

Before generating any claim-bearing outcome, author and stabilize every program, register the decisive observable and decision rule, and freeze the complete execution pipeline. Put the simulator in execution_binding and every downstream capability program in additional_execution_bindings; enumerate every exact scientific argv and commission every distinct program under its own instrument claim with representation, physics_controls, boundaries, diagnostics, and numerical_regime checks. Every capability name must be selected exactly from the installed registry. Plain run_python is workbench-only and cannot qualify a downstream scientific evidence stage. Do not repair a scientific contract after outcomes and then close the same claim decisively; if the protocol must change, close nondecisively or create a successor and collect fresh evidence.

Bind the decisive JSON observable to machine-readable estimator or formula, component and sign convention, units, normalization, time or window rule, seed pairing, and uncertainty rule. Select exact header-named nonoverlapping field and particle energy totals. Continue autonomously without asking the operator to choose physics or parameters. Stop with a bounded supported, falsified, or honestly unresolved conclusion and account explicitly for numerical fidelity, sampling uncertainty, prior non-evidentiary discovery context, and every open or closed claim.
