Skills for AI Agents

mlmm-toolkit ships instructions for AI agents (Claude Code, Codex, Cursor, and others) in the skills/ directory. With the skills loaded, an agent can run and analyze calculations from plain-language instructions.

First example

To install them, tell your AI agent:

Import https://github.com/t-0hmura/mlmm_toolkit/tree/main/skills as skills, and install mlmm-toolkit by following mlmm-install.

If you cloned the GitHub repository, you can give the local skills/ path instead. Then you can ask, for example:

Read the paper, build a model from the PDB structure PDB ID, and study the mechanism of the reaction step with the mlmm-toolkit skills.

Skills

Each folder in skills/ is one skill (<name>/SKILL.md).

Skill

What it covers

mlmm-overview (start here)

What mlmm-toolkit is, which of the three all modes fits your structures, TS strategy, and how to read the outputs

mlmm-cli

When to use each subcommand, how to judge success, and pitfalls and recovery

mlmm-model-setup

Structure files and Amber parm7/rst7, the charge and multiplicity, the ML region and layers, and keeping the same atoms across states and variants

mlmm-install

Installing mlmm-toolkit, the MLIP backends, DFT, xTB, and AmberTools, and probing an unknown environment

mlmm-hpc

PBS / SLURM job scripts, walltime, monitoring, and dynamic dispatch

colab-local-gpu-runtime

Running the Colab interface on a local NVIDIA GPU through WSL2 and Docker Desktop on Windows

Install by hand

Copy the skill folders into your agent’s skill directory:

  • Claude Code: .claude/skills/ in a project, or ~/.claude/skills/ for all projects.

  • Codex: .agents/skills/ in a repository, or ~/.agents/skills/ for all repositories.

For example, from the root of the repository:

mkdir -p ~/.claude/skills
cp -r skills/mlmm-* skills/colab-local-gpu-runtime ~/.claude/skills/

The skills leave exact flags and defaults to the installed CLI (mlmm <subcommand> --help-advanced) and the command reference.

See also