research-workflow-dev

Domain-agnostic research-coding workflow discipline for computational science (JAX/Python research family). 70 skills across ideate / literature / scope / build-correctly / equation-critical sources / verify / inference-rigor / review / performance-and-scale / record / communicate / reproduce-and-release, plus an equation-verifier agent and six slash commands. v1.4.x–1.5.0 adds a true front-of-funnel (research ideation + brainstorming + prior-art + literature discipline), Bayesian/statistical inference rigor (MCMC convergence, predictive checks, model selection), HPC performance & scale (profiling, scaling, JAX performance, cluster-run contract), software/data release & citation (CITATION.cff/DOI, release checklist, data management plans), and a figure craft & interpretation layer (expert astrophysics plotting in a house style, plot-craft review, and figure-interpretation guarding). The suite covers evidence-first execution, numerical validation, reproducibility, PDF equation extraction, equation-to-code traceability, reference-license firewalls, computational-physics code review, figure faithfulness, MyST documentation authoring, eight self-limiting enforcement hooks, and per-domain lenses for reference-parity work.