surajinacademia/computational-modeling-skills
Fourteen skills for research project setup, scientific modeling, paper reproduction, model documentation, Jupyter notebooks, linear stability analysis, quantitative plots, scientific schematics, microscopy image analysis, scientific claim review, staged scientific writing, request clarification, HPC Slurm jobs and sweeps, and end-of-session research lessons.
Critically review and critique scientific claims against their actual results, calculations, simulations, figures, controls, and literature so claims are rigorous and bounded in scope. Use to review a researcher's own claims, review a published paper or preprint, or organize reviewed claims into a working story arc. Do not ghostwrite manuscripts, build paragraph-level outlines, invent evidence, or infer novelty or significance without domain literature.
Formulate, implement, verify, or review physics-based scientific models and theoretical derivations. Use for model development, hypothesis tests, and physical validation; exclude mechanism-free empirical fitting and surrogate or machine-learned models.
Create, style, review, or simplify quantitative plots from completed scientific data, including notebook figures and ordinary paper-ready PDFs. Use the visual gallery, a portable style adapter, and direct plotting templates while preserving scientific identity, weighting, and observable definitions. Do not use for simulation execution, generic data cleaning, inferential modeling, spatial simulation frame or video rendering, or composite schematic design.
Prepare, submit, expand, monitor, cancel, and recover HPC Slurm jobs and independent parameter sweeps using native scheduler commands. Use live account and QOS evidence to fill eligible idle capacity without duplicate work; preserve established project launchers and scientific settings.
Analyze experimental microscopy images, volumes, and time series for measurements such as cell counts, fluorescence, morphology, tracking, and matrix deformation. Use one suitable image-analysis MCP server by default and produce a matched before-and-after figure and reproducible supplementary methods for each completed analysis.
Create, maintain, execute, or review Jupyter notebooks for scientific research, theoretical modeling, simulations, and scientific data analysis. Use for new research notebooks and targeted edits to existing .ipynb files; standalone model documents and business dashboards are outside this skill's scope.
Clarify vague or incomplete requests by spotting holes in user intent, supplying the facts and tradeoffs needed for informed decisions, confirming unresolved choices, and completing the task. Use when invoked, when checking shared understanding, when a request has ambiguity or missing specifications that change the work, when the user asks for help deciding, or when another workflow delegates an unresolved brief, plan, or scope choice. Do not activate for ordinary retrieval, definitions, or already-precise requests.
Derive, compute, plot, and report physical linear stability analyses for stationary states of 1D or 2D continuum systems. Use for reaction-diffusion, mechanical, curvature, and related spatial dynamical equations requiring analytical dispersion relations, numerical eigenspectra, or a reproducible LaTeX report. Exclude nonlinear time integration and stability analysis of numerical time-stepping schemes.
Plan, write, revise, or audit one scientific model document from code or an approved plan. Explain physics, equations, and methods; record MIASE simulation information, model sources, and change history. Use when a model document is the deliverable, not for isolated explanations.
Reproduce, assess, or apply a scientific paper's computational model, method, claims, or results. Use for paper-to-code reconstruction, model verification, hypothesis testing, comparison with published experimental or computational evidence, or a requested adaptation. Do not use for a literature summary without an implementation or assessment target.
Set up or reorganize a research repository using the maintainer's template, Good Research Code practices, and a Git workflow for verified checkpoints, concurrent work, and recovery. Use only when the user explicitly requests research-project-setup or explicitly asks to apply this template. Routine repository opening, coding, error diagnosis, and general maintenance do not activate it.
End-of-chat review of a scientific problem-solving chat that recaps the problem and solution, records rules for the agent in agent_lessons.md, and records learning topics and lessons for the user in my_lessons.md. Use only when the user types the /ruminations command ($ruminations in Codex). Do not run on the word "recap", other requests to summarize, or unprompted at the end of a chat.
Generate or reconstruct scientifically faithful figures from specifications, equations, data, or image references. Use for composite publication design, scientific schematics, image reconstruction, or exact PDF and outlined-SVG export requirements. Do not use for an ordinary exploratory, diagnostic, comparative, or single scientific data plot; data-visualization handles those plots.
Help a researcher turn notes, figures, a story arc, an outline, or a partial manuscript into a scientific paper through staged writing. Use to choose what to write next, build or revise the paragraph-level outline, draft a section from supplied findings, or revise a draft whose structure or length needs work.