メニュー
Use when auditing CFD-AI/SciML papers for reproducibility: governing equations, nondimensional numbers, geometry, mesh, boundary conditions, solver settings, datasets, splits, baselines, hyperparameters, compute, and exact commands.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | cfd-reproducibility-checker |
| description | Use when auditing CFD-AI/SciML papers for reproducibility: governing equations, nondimensional numbers, geometry, mesh, boundary conditions, solver settings, datasets, splits, baselines, hyperparameters, compute, and exact commands. |
Use when a manuscript reports CFD simulations, CFD-generated datasets, neural surrogates, PINNs, closure models, or flow-control experiments and needs reproducibility review.
rubrics/cfd-reproducibility-rubric.md before scoring.references/gold-papers/fukami-2019-super-resolution-jfm.md.references/gold-papers/fukami-2021-spatiotemporal-super-resolution-jfm.md.references/gold-papers/lee-2019-cylinder-wake-jfm.md.references/gold-papers/maulik-2020-probabilistic-neural-networks-prf.md.references/gold-papers/vinuesa-2023-drl-drag-reduction-epje.md.| Task | Extra reproducibility fields |
|---|---|
| Spatio-temporal SR | Spatial downsampling factor, temporal frame interval, input/output variables, reference-resolution source. |
| Wake prediction | Re split, domain, BC/IC, solver, mesh, timestep/CFL, rollout horizon, force/frequency diagnostics. |
| Neural ROM | Full-order solver, latent dimension, rollout horizon, POD/Galerkin baseline setup. |
| DRL/control | Simulator version, action limits, observation locations/variables, reward, baseline controller, seeds/training budget. |
| Item | Present? | Evidence location | Risk | Required fix |
|---|
Also include:
| Rubric axis | Score 0-3 | Evidence | Blocking TODO |
|---|
Use when writing, reviewing, auditing, revising, or producing LaTeX for CFD-AI and scientific-ML manuscripts, reviews, experiment plans, reproducibility audits, related-work positioning, reviewer responses, or benchmark evaluations.
Use when stress-testing CFD, fluid mechanics, turbulence, PINN, neural-operator, surrogate, closure, or SciML manuscripts from a strict reviewer perspective.
Use when designing figures, result flow, captions, and visual evidence for CFD-AI/SciML journal manuscripts.
Use when coordinating multiple specialized CFD-AI/SciML paper agents: reviewer, evidence auditor, experiment planner, figure editor, prose editor, and final gatekeeper.
Use when auditing a CFD-AI/SciML manuscript's abstract, introduction, contributions, or discussion to verify that every claim is supported by evidence and scoped correctly.
Use when running iterative reviewer-to-editor revision of a CFD-AI/SciML manuscript: reviewer audit, editorial rewrite, re-audit, residual-risk tracking, and submission-readiness gating.