nep-gpumd

name	nep-gpumd
description	Route NEP requests to task-specific subskills. NEP (Neuroevolution Potential) is the native machine-learning potential family of the GPUMD ecosystem — analogous to DeePMD-kit for LAMMPS. Use when the user asks for `nep.in`, `train.xyz`, `test.xyz`, NEP training, NEP89 reuse, prediction mode, fine-tuning, dipole / polarizability auxiliary models, or automation via NepTrain / NepTrainKit.
compatibility	Requires a NEP-capable GPUMD build providing the `nep` executable (usually shipped together with `gpumd` in modern releases). Training needs labeled datasets in NEP-compatible extxyz format.
license	GPL-3.0-only
metadata	{"author":"Jhin","version":"0.2.0","repository":"https://github.com/brucefan1983/GPUMD","documentation":"https://gpumd.org/nep/"}

NEP Task Router

Use this skill as the top-level routing layer for NEP work in the GPUMD ecosystem. NEP is tightly coupled to GPUMD: a trained nep.txt file is loaded by gpumd through potential nep.txt, and the labeled datasets used for training are concatenated extxyz files just like GPUMD's model.xyz.

If the task is about running MD with an existing NEP potential rather than training one, route to molecular-dynamics/gpumd instead.