Matlantis

Matlantis Agent Skills

Japanese version / 日本語版

A plugin marketplace providing Matlantis / PFP atomistic simulation skills for Claude Code and Codex.

Overview

Matlantis Agent Skills is a collection of skills that enables Claude Code and Codex to assist with atomistic simulations on the Matlantis platform. 16 skill modules guide Python code generation for structure optimization, molecular dynamics, property calculations, and more — all powered by PFP (Preferred Potential) neural network potentials.

Each skill systematically covers implementation patterns, best practices, and common errors with solutions, all built on ASE (Atomic Simulation Environment).

Skills Catalog

Category	Skill	Summary
Foundation	matlantis	Platform overview, PFP calculation mode selection, library architecture
Foundation	setup	Calculator initialization, calc_mode configuration, structure file I/O
Foundation	ase-basics	ASE fundamentals, Atoms object, molecule & crystal construction
Structure Building	modeling	Surface slab generation, SMILES→3D conversion, substitution/doping, liquids & amorphous
Structure Building	external-data	Structure retrieval from PubChem, Materials Project, etc.; file format conversion
Computation	optimization	Fixed-cell & variable-cell relaxation, symmetry preservation, multi-stage optimization
Computation	dynamics	NVE/NVT/NPT molecular dynamics, metadynamics, transport properties
Computation	reaction	NEB reaction path search, transition state search, activation energy
Computation	adsorption	Adsorption structure construction, adsorption energy, adsorption site search
Computation	property-analysis	Elastic constants, phonons, QHA, diffusion coefficients, viscosity, thermal conductivity
Workflow	integrated-workflows	Structure screening, adsorption campaigns, defect analysis pipelines
Workflow	background-job	Background execution of long-running calculations, checkpoint management
Acceleration	light-pfp	Light-PFP custom models (15–40x faster) training & inference
Utility	pfcc-extras	Visualization, structure processing, and job execution control utilities
Utility	visualization	Atomic structure & trajectory visualization, nglview, POV-Ray
Infrastructure	storage	Group Drive persistent storage, team data sharing
Infrastructure	ssh	SSH connection setup, VSCode Remote-SSH, WinSCP integration

Getting Started

Prerequisites

• Claude Code CLI installed when using Claude Code
  • A version supporting the plugin feature is required
  • Installation: https://claude.ai/code
• Codex CLI installed when using Codex plugins
  • Installation: https://developers.openai.com/codex/cli

Installation

Claude Code

Run the following commands in the Claude Code REPL:

# 1. Register the marketplace
/plugin marketplace add https://github.com/matlantis/matlantis-agent-skills

# 2. Open the plugin manager
/plugin

# 3. Enable the matlantis plugin from the "Marketplaces" tab

Codex

Run the following commands from your shell:

# 1. Register the marketplace
codex plugin marketplace add matlantis/matlantis-agent-skills

# 2. Install the plugin from the marketplace
codex plugin add matlantis@matlantis-agent-skills

Once installed or enabled, the agent will automatically reference the appropriate skills when you work with Matlantis-related code or questions.

Skills Details and Usage Examples

matlantis

Platform overview skill. Covers PFP calculation mode selection guidelines (PBE, R2SCAN, WB97XD, etc.), library hierarchy, and critical constraints.

"I want to calculate the surface energy of copper. Which calc_mode should I use?"

setup

Calculator initialization and environment setup. Covers Estimator/ASECalculator configuration patterns, structure file I/O, and batch execution settings.

"Write code to initialize the PFP Calculator in R2SCAN mode"

ase-basics

ASE fundamentals. Covers Atoms object creation/editing, molecule & crystal construction, periodic boundary conditions, and file I/O.

"Load a crystal structure from a CIF file and create a 2x2x2 supercell"

modeling

Structure modeling. Covers surface slab generation (ASE builders / pymatgen), 3D molecule generation from SMILES (RDKit), atomic substitution/doping, and construction of liquids, amorphous materials, and high-entropy alloys.

"Create a 4-layer FCC copper (111) surface slab with 15 Å of vacuum"

external-data

Structure retrieval from external databases. Covers integration with PubChem (organic molecules), Materials Project (inorganic crystals), AFLOW, COD, and file format conversion.

"Retrieve the BaTiO3 crystal structure from Materials Project and optimize it"

optimization

Structure optimization (relaxation). Covers fixed-cell optimization (atomic positions only) and variable-cell optimization (including lattice vectors), LBFGS/BFGS/FIRE optimizers, and symmetry-preserving optimization.

"Relax the crystal structure with variable-cell optimization, preserving symmetry"

dynamics

Molecular dynamics (MD) simulation. Covers NVE/NVT/NPT ensembles, Langevin thermostat, PLUMED metadynamics, trajectory analysis, and transport property calculations (diffusion, viscosity, thermal conductivity).

"Run a 100 ps NVT MD simulation at 300 K"

reaction

Reaction path and transition state search. Covers NEB (Nudged Elastic Band), CI-NEB, and reaction barrier calculations using ReactionStringFeature.

"Calculate the activation energy for CO dissociation on a surface using NEB"

adsorption

Molecular adsorption on surfaces. Covers adsorption structure construction with add_adsorbate, adsorption energy calculation with D3 dispersion correction, and adsorption site search (Optuna black-box optimization recommended).

"Find the most stable adsorption site for CO on Pt(111) and compute the adsorption energy"

property-analysis

Property calculations and MD post-processing. Covers elastic constants/moduli, phonon band structures & density of states, quasi-harmonic approximation (QHA), diffusion coefficients, viscosity, thermal conductivity, and radial distribution functions (RDF).

"Calculate the phonon band structure and density of states for silicon"

integrated-workflows

Integrated workflow (pipeline) templates. Provides structure screening (multi-structure batch optimization + energy ranking), adsorption campaigns (site search + ranking), and defect analysis pipelines.

"Batch-screen 10 alloy compositions and rank them by stability"

background-job

Background execution of long-running calculations. Covers detached computation from notebook sessions, checkpoint/restart, and job monitoring.

"Run a large-scale MD in the background and monitor progress"

light-pfp

Light-PFP custom models. Covers the workflow for training MTP (Moment Tensor Potential) models specialized for specific materials, achieving 15–40x faster inference than PFP.

"Create a Light-PFP model specialized for copper-based materials"

pfcc-extras

pfcc_extras utility toolkit. Covers the visualization layer (show_gui, view_ngl), structure processing layer (partial occupancy, molecule wrapping), MD event scheduling, and job execution control.

"Run multiple structure optimization jobs in parallel with resource limits"

visualization

Atomic structure & trajectory visualization. Covers simple display with pfcc_extras.show_gui, interactive visualization with nglview, trajectory animation, and high-quality POV-Ray rendering.

"Display an MD trajectory as an animation and render specific frames in high quality"

storage

Group Drive persistent storage. Covers file upload/download to cloud object storage and team data sharing.

"Save calculation results to Group Drive and share with team members"

ssh

SSH connection setup. Covers WebSocket-based SSH tunneling, VSCode Remote-SSH integration, and file transfer via WinSCP.

"Connect to the Matlantis environment from VSCode via Remote-SSH"

License

Copyright 2026 Matlantis Corporation.

This project is licensed under the Apache License 2.0.

Third-Party Licenses

This repository contains documentation from Matlantis, which is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0). Copyright 2021-present, Preferred Networks, Inc. and ENEOS Corporation.