Design scientifically defensible computational experiments and reviewable VASP calculation plans before task preparation. Use when converting research questions into falsifiable hypotheses, observables, controls, convergence studies, validation checks, uncertainty targets, calculation matrices, resource budgets, stopping rules, or approved design revisions. Do not generate VASP inputs, submit jobs, analyze completed outputs, or archive tasks; hand those stages to vasp-workflow, vasp-analysis, and vasp-work-manager.
Interview, challenge, plan, preview, and initialize scientific research projects with a reproducible docs/code/raw_data/formal_data structure. Use when starting a new research project, converting an early idea into testable hypotheses and executable experimental or computational objectives, standardizing an existing project directory without overwriting files, or establishing a validated data lifecycle for publication-ready results.
Design the research narrative for a scientific presentation before any slide-level work begins. Use when converting research materials, papers, notebooks, experiment results, or calculation outputs into a structured talk plan with a core argument, evidence chain, audience profile, narrative arc, time budget, material priorities, and predicted questions. Output a talk_plan.md that hands off to simple-sci-ppt for slide-level planning and PPTX generation. Do not create slides, write generator code, select layout archetypes, or produce PPTX files; hand those to simple-sci-ppt.
Create, revise, and verify formal academic PowerPoint decks with formula/image/table/code cards, section tabs, and semantic containers for research reports, lectures, course reviews, exercise sessions, thesis/group meetings, and single-topic scientific slides. Use when Codex is asked to use simple-sci-ppt, simple_sci_ppt, or simple_sci_ppt_gen, or to turn PDFs, PPTX files, notes, papers, notebooks, figures, or reference folders into a planned PPTX with Markdown planning, MathJax formulas, editable `pptxgenjs` objects, speaker notes, and PowerPoint validation. This skill owns slide-level planning and PPTX generation. For narrative design, use `sci-talk-planning` first and pass its `talk_plan.md`.
Process and analyze completed VASP outputs by extracting validated plot-ready .dat files, generating figures, evaluating approved observables, writing interpretation reports, and proposing versioned computation-design changes without submitting jobs. Use when converting VASP, phonopy, LOBSTER, PyProcar, VASPKIT, CHGCAR/ELFCAR/PARCHG, DOSCAR, EIGENVAL, PROCAR, band.yaml, total_dos.dat, COHPCAR, ICOHPLIST, or notebook outputs into reusable data products. Do not create tasks or archive ledgers; hand design revisions, execution, and records to computation-design, vasp-workflow, and vasp-work-manager.
Records/archive layer for existing VASP calculation work: register, record, archive, report, and verify tasks without creating, submitting, recovering, or changing calculations. Use when preserving computation-design snapshots/approvals, workflow tasksets, state.json, task_spec.json, submission reviews, queue logs, core outputs, .dat data, figures, manifests, SHA256 checksums, archive versions, or task history for nmg/Phoenix/G3 work. Do not perform scientific design, VASP input generation, Slurm execution, recovery, or analysis; hand those to computation-design, vasp-workflow, and vasp-analysis.
Running/orchestration layer for VASP calculation tasks after scientific design: structure/POSCAR research, input preparation, submit review, safe Slurm submission, queue/dependency handoff, parsing, bounded recovery, and automation. Use when preparing approved or exploratory test/relax/SCF/band/DOS/phonon tasks, verifying computation-design provenance, confirming POSCAR/INCAR/KPOINTS/POTCAR/job.sh inputs, selecting nmg/Phoenix/G3 resources, submitting jobs, or handling runtime failures. Do not own scientific experiment design, archive ledgers, or plot/data analysis; hand those to computation-design, vasp-work-manager, and vasp-analysis.