Guide · Agent skill
How to use the Molecular Dynamics Simulation agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The molecular-dynamics skill gives an AI agent a tested workflow for running and analyzing molecular dynamics (MD) simulations, which model how atoms in a protein or small molecule move over time. It is written for structural biologists, biophysicists, computational chemists and Python developers who already understand the basic science. By the end of this guide you will know what the skill does, how to install it, how to ask for help with natural requests, and what to check before trusting the results.
Molecular Dynamics SimulationK-Dense-AI/scientific-agent-skills · MIT Download the skill Quick start
- 1Download the skill. Download the molecular-dynamics skill ZIP from this page.
- 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder for one project only.
- 3Or upload it in the Claude apps. Alternatively, upload the ZIP under Settings → Capabilities → Skills.
- 4Create the Python environment. Run the uv commands in this guide to create a Python 3.12 environment with OpenMM, MDAnalysis, matplotlib and pandas.
- 5Ask for a specific task. Describe your structure file, your scientific question and the analysis you want, such as RMSD or a contact map.
- 6Review results before using them. Check the preparation choices, the logs and the minimized structure before interpreting any trajectory analysis.
What does the molecular-dynamics skill make the agent do?
With this skill installed, the agent can help you prepare a molecular structure, add water and ions, run energy minimization and equilibration, run production simulations with OpenMM, and analyze the resulting trajectories with MDAnalysis.
Analysis examples include RMSD, RMSF, contact maps and free energy surfaces. The skill covers protein and small molecule systems, force field definitions, and use cases in structural biology, drug binding and biophysics.
- System preparation from a PDB file, with hydrogens, solvent and ions added
- Energy minimization to remove steric clashes
- NVT equilibration and production MD with OpenMM
- Trajectory analysis with MDAnalysis (RMSD, RMSF, contact maps, free energy surfaces)
How do you install the molecular-dynamics skill?
Download the skill ZIP from this page. Then add it to Claude in one of two ways: unzip it into ~/.claude/skills/ for your user account, or unzip it into a project's .claude/skills/ folder to make it available only in that project. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
Agent skills are folders containing a SKILL.md file with instructions, plus supporting reference files. The skill's files are SKILL.md, references/system_preparation.md and references/mdanalysis_analysis.md.
- Requirement: Python 3.11 or newer
- The examples in the skill use Python 3.12 through uv
- Network access is needed for installation only; simulation and analysis run locally
How do you set up the Python environment for OpenMM and MDAnalysis?
The skill recommends a dedicated environment so the simulation packages do not clash with your other Python tools. The command below creates one with uv and installs OpenMM, MDAnalysis, matplotlib and pandas.
On Windows, use the Python path shown in the second line of the command block instead. Optional tools PDBFixer and OpenFF must be installed separately. PDBFixer can be installed in a separate environment with conda if needed.
- PDBFixer and OpenFF are optional and are not installed by the main command
- Check that the environment works before starting a long simulation
uv venv --python 3.12 .venv-md
uv pip install --python .venv-md/bin/python "openmm==8.6.1" "MDAnalysis==2.10.0" matplotlib pandas
# Windows: use .venv-md/Scripts/python.exe instead
# Optional PDBFixer, in a separate environment:
conda install -c conda-forge pdbfixerHow do you trigger the skill with natural requests?
You do not need special syntax. Describe the scientific goal in plain language and mention the structure file or trajectory you have. The skill is meant to be triggered when your request involves MD setup, simulation or trajectory analysis.
Good requests name the molecule, the question and the files. Vague requests such as 'simulate my protein' are less useful than requests that state what you want to learn.
- 'Prepare cleaned_protein.pdb for an OpenMM run with 10 Å of water padding and 0.15 M salt.'
- 'Minimize and equilibrate this system, then save an NVT log and DCD trajectory.'
- 'Compute RMSD and RMSF for the backbone in trajectory.dcd against the starting structure.'
- 'Make a contact map between the ligand and the binding pocket residues from my trajectory.'
What does a typical workflow look like?
A typical run moves through preparation, minimization, equilibration, production and analysis. Each stage should write to its own output prefix so earlier results are not overwritten. The skill's examples use 2 fs time steps, so 50,000 NVT steps equal 100 ps.
Before parameterization, review the biological assembly, alternate locations, missing loops, termini, protonation, disulfides, ligands and cofactors. The skill's notes say a trajectory alone does not establish binding free energy or residence time.
- Use PDBFixer or a cleaned PDB file as the starting structure
- Use a new output prefix for each stage
- Check atom order, molecule integrity, box vectors and time units in MDAnalysis before analysis
What should you know about safety, scripts and network use?
The skill runs Python scripts on your machine. Its stated network use is for installation only, and simulation and analysis run locally. Review any script before running it, and run it in the dedicated environment rather than your main Python setup.
The skill's notes say that reference-platform checks use small synthetic systems. They do not establish GPU performance or scientific convergence. A registered GPU plugin does not prove that a device and driver work, so test your chosen platform first.
- Runs scripts: yes
- Uses network: no during simulation and analysis
- Outputs are exploratory and must not guide clinical or medical decisions
- Long production runs on a laptop without GPU support can take impractically long
What tips will help you get reliable results?
Start with a small system and a short run to confirm the pipeline works. Then scale up. Keep the minimized structure and the log files, because they show whether the system settled before production.
Be careful with preparation choices. PDBFixer can propose missing residues and replace modified residues, but neither step establishes the correct experimental construct. The skill's notes say its removeHeterogens setting keeps water and removes ligands, ions and cofactors, so inspect any ligands and parameterize them separately.
- Test on a small system before a long run
- Keep logs and minimized structures for each stage
- Parameterize ligands and nonstandard residues separately
- Describe your scientific question clearly so the agent can pick the right analysis
What people use it for
- Testing how a mutation changes the flexibility of a protein by comparing RMSF across trajectories
- Setting up a protein in water with ions and running energy minimization before a production run
- Measuring how far a protein backbone drifts from its starting structure using RMSD
- Building a contact map between a ligand and nearby residues to see which contacts persist during a trajectory
Frequently asked questions
How do I install the molecular dynamics agent skill?
Download the skill ZIP from this page and unzip it into ~/.claude/skills/ for your user account, or into a project's .claude/skills/ folder. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
What do I need to run the molecular-dynamics skill?
You need Python 3.11 or newer, with OpenMM and MDAnalysis installed, plus matplotlib and pandas. PDBFixer and OpenFF are optional and must be installed separately. Network access is needed for installation only.
Is the molecular-dynamics skill free to use?
The skill is released under the MIT licence. The source does not list prices, so check the licence file in the skill's repository for its terms.
Does the skill need a GPU?
The skill's notes do not establish GPU performance. OpenMM can use GPU support, but a registered GPU plugin does not prove that a working device or driver is available. Long production runs on a laptop without GPU support can take impractically long.
How does this compare with other MD tools?
The skill is built around OpenMM for simulation and MDAnalysis for trajectory analysis. This guide does not compare it with other tools, so choose based on whether you need these two libraries and a Python workflow.
Can I use simulation results for clinical decisions?
No. The skill's outputs are exploratory and do not establish real-world binding or efficacy, so they must not be used to guide treatment. A trajectory alone also does not establish binding free energy or residence time.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are a structural biologist, biophysicist or Python developer who wants an agent to help set up and analyze OpenMM and MDAnalysis simulations, and you can check the preparation choices yourself. Pick something else if you need clinical or medical conclusions, if you cannot install Python 3.11 or newer with these libraries, or if you plan long production runs without GPU access.
Download the skill