Die Oberfläche ist lokalisiert; Skill-Beschreibungen und -Dokumente werden per KI automatisch in Ihre Sprache übersetzt. Code, Befehle, Pfade und Fachbegriffe bleiben im Original, maßgeblich ist das Dokument des Autors.
Skills entdecken
Daten und Forschung · SKILL.md

lammps-evidence-md

lammps-evidence-md

Evidence-first LAMMPS molecular-dynamics workflow for input decks, potentials, minimization, equilibration, deformation, thermo logs, trajectories, restart/data files, stress-strain extraction, and reproducibility checks. Use when Codex must prepare, run, diagnose, or report a LAMMPS materials simulation or hand its trajectory to OVITO without overstating n…

Independent GitHub snapshot · Automated classification · Original documentation governs

Skill-Dokumentation

GitHub

name: lammps-evidence-md description: Evidence-first LAMMPS molecular-dynamics workflow for input decks, potentials, minimization, equilibration, deformation, thermo logs, trajectories, restart/data files, stress-strain extraction, and reproducibility checks. Use when Codex must prepare, run, diagnose, or report a LAMMPS materials simulation or hand its trajectory to OVITO without overstating nanoscale demo results as calibrated macroscopic properties.

LAMMPS Evidence MD

Keep the input deck, potential provenance, random seeds, executable build, log, trajectory, and postprocessing chain traceable.

Start

ai-cae-toolbox context-scope --solver lammps
ai-cae-toolbox bridge-plan --solver lammps --objective "<objective>"
ai-cae-toolbox create-run --solver lammps --case <case-name> --objective "<objective>"
ai-cae-toolbox write-smoke-template --solver lammps --run-dir <run-dir>

Use lammps_check_installation and lammps_run_input_file when MCP is available.

Build and Run

  • Declare units, boundary conditions, atom style, dimensions, masses, and atom types.
  • Record potential filename, source, license, supported species, and parameterization scope.
  • Separate minimization, equilibration, loading, and production phases.
  • Record timestep, thermostat/barostat damping, deformation rate, dump interval, thermo fields, random seeds, and restart strategy.
  • Run a tiny parser/smoke case before a long trajectory.

Read references/validation.md before interpreting stress, fracture, diffusion, or phase behavior.

Finish

Require a normal LAMMPS termination signal, complete log, expected step count, thermo data, and trajectory or final data output. Export derived curves through a versioned script. Hand trajectories to the OVITO skill for visualization, but keep LAMMPS solve evidence and OVITO render evidence separate.

Gespeicherte Skills