Computational Chemistry Agent Skills
Skills for computational chemistry: ASE, DFT codes, LAMMPS, DeePMD, RDKit, Open Babel and more, built for OpenClaw
- 157 ★ on the repository
- Last commit
- LGPL-3.0
Extended access
Why this level
- The skills run calculation programs and scripts on a compute node and write result files
- To install, the agent downloads an archive from GitHub and runs `npx` for each folder
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/computational-chemistry-agent-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skills run calculation programs and scripts on a compute node and write result files
- To install, the agent downloads an archive from GitHub and runs `npx` for each folder
This is third-party code. Review the repository files before installing.
What it does
Computational Chemistry Agent Skills is a repository from the Jinzhe Zeng group with agent skills (following the agentskills.io specification) for computational chemistry. Automation runs through OpenClaw but the format is shared. Skills are grouped by task: atomistic workflows (ASE with GPAW and MACE calculators, relaxation, static, dynamics and NEB scenarios), quantum chemistry (DFT in ABINIT, CP2K and GPAW with subfolders for bands, density of states, dynamics and relaxation), machine learning potentials (DeePMD: training, DPA3 fine-tuning, inference, DP-GEN), molecular dynamics (LAMMPS with DeePMD and ReaxFF, antechamber), data processing (dpdata, Open Babel, Packmol, pymatgen), molecular representation and conformers (RDKit, Uni-Mol), analysis (phonopy, ReacNetGenerator) and an agent task board manifest. A list of all skills with descriptions is on the site skills.jinzhezeng.group. The work is accompanied by a paper in J. Chem. Theory Comput. (2026), "Automating Computational Chemistry Workflows via OpenClaw and Domain-Specific Skills", and the authors ask to cite it when you use the repository. For installation there is a ready phrase for an OpenClaw agent that downloads the archive and installs all visible skills with `npx -y skills add <skill folder> -a openclaw -y`.
Who it is for. Computational chemists and materials scientists who run calculations in ABINIT, CP2K, LAMMPS or DeePMD and want to hand routine work to an agent.
Good fit when
- You need to prepare and run a relaxation or a static calculation in ABINIT, CP2K or GPAW
- You need to train or fine-tune a DeePMD neural potential and then run dynamics in LAMMPS
- Structures have to be converted between formats and a molecule mixture built with Packmol and Open Babel
- You need conformers and molecular descriptors in RDKit or Uni-Mol
Not a fit when
- You need the calculation packages themselves: the skills only describe running them and ABINIT, CP2K, LAMMPS and others must be installed on a server
- You need general chemistry help without calculations: the skills target computational packages
- You need to work outside OpenClaw without adaptation: the author tested that environment
Example request
Prepare a CP2K structure relaxation from my data and run it on the node, then show the final energyLimitations
The authors tested the skills in OpenClaw, and in Claude Code or Codex you adapt and verify them yourself. They do not contain the calculation programs themselves, so packages and their licenses (for example commercial codes) must be set up separately. Calculations can be heavy and long and are run on compute nodes, not in chat.
The repository is under the LGPL-3.0 license. For publications the authors ask to cite their paper. The skills refer to formats and parameters of specific program versions, so check results against the package documentation.
How to disable. Delete the installed skill folders from the OpenClaw (or other agent) directory. Calculation work folders and results are stored separately and the skill does not touch them.
FAQ
Which calculation packages are covered?
ABINIT, CP2K, GPAW, ASE, LAMMPS, DeePMD, DP-GEN, RDKit, Open Babel, Packmol, pymatgen, phonopy, ReacNetGenerator, antechamber, Uni-Mol and dpdata. The full list is on the project site.
How do I install the skills?
Send the OpenClaw agent the phrase from the README: it downloads the repository archive and runs `npx -y skills add <folder> -a openclaw -y` for each folder.
Do the skills work in Claude Code?
The agentskills.io format is shared but the authors tested OpenClaw. In Claude Code copy the needed folders into the skills directory and verify them on test calculations.
Do I need to install the programs themselves?
Yes, the skills describe launching and input preparation but ABINIT, CP2K, LAMMPS and the rest must be installed on a compute node.
How do I cite it?
The authors ask to cite the Ding et al. paper in J. Chem. Theory Comput. 2026, DOI 10.1021/acs.jctc.6c00622.
Discussion
No questions or reviews yet.
Related
Skills37 skills for computational drug discovery: from target selection and docking to ADMET and the clinical record.
Around 200 skills for Claude Code on bioinformatics, chemistry and biostatistics, from RNA-seq analysis to molecular docking
17 skills for computational biology: RNA-seq, single-cell, protein structure, docking, literature and figures
Plugin with 130-plus skills for atomistic materials research: relaxation, dynamics, DFT and structure databases through MCP.
24 skills for numerical methods and materials science: the agent checks stability, meshes, convergence and cluster runs.
166 skills for scientific work: bioinformatics, cheminformatics, clinical data, geospatial analysis and 100+ databases
README in short
The repository publishes a set of skills that let an agent set up and run computational chemistry tasks in OpenClaw. The description comes with a scientific publication and a site with a skill catalog. The code has a schema for validating skill metadata and scripts that generate the skills table.
README badge
Original SKILL.md
--- name: create-skill description: Create new Agent Skills following the agentskills.io specification. Use when the user wants to create, scaffold, or design a new skill for AI agents. Handles SKILL.md generation, directory structure setup, and validation. license: MIT metadata: author: njzjz-bot version: '1.0' --- # Create Skill This skill helps you create new Agent Skills that follow the [agentskills.io specification](https://agentskills.io/specification). ## Quick Start When asked to create a new skill: 1. **Gather requirements**: Ask what the skill should do 1. **Choose a name**: lowercase letters, numbers, hyphens only (e.g., `pdf-processing`, `data-analysis`) 1. **Generate the structure**: Create `SKILL.md` with proper frontmatter 1. **Add optional components**: scripts, references, assets as needed ## Directory Structure ``` skill-name/ ├── SKILL.md # Required: main skill file ├── scripts/ # Optional: executable code ├── references/ # Optional: additional documentation └── assets/ # Optional: static resources ``` ## SKILL.md Template ```markdown --- name: your-skill-name description: What this skill does and when to use it. Be specific and include keywords that help agents identify relevant tasks. Max 1024 characters. license: MIT compatibility: Optional - environment requirements if any metadata: author: your-name version: "1.0" allowed-tools: Optional - pre-approved tools (experimental) ---