Skip to content
Get packs

Guide · Agent skill

How to use the RDKit agent skill: install, trigger and examples

Updated · written with AI from the project's own documentation and checked against our catalogue data

The RDKit skill teaches an AI agent to work with chemical structures using RDKit, an open-source cheminformatics library. Once installed, you can ask the agent to read molecules from SMILES or SDF, calculate properties, run searches and script workflows. This guide explains what the skill does, how to install it, how to trigger it, and what to watch for.

RDKit Cheminformatics ToolkitK-Dense-AI/scientific-agent-skills · MIT

Quick start

  1. 1Download the ZIP. Get the rdkit skill ZIP from the page where you found this guide.
  2. 2Unzip the skill. Extract it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
  3. 3Or upload it in the Claude apps. Go to Settings → Capabilities → Skills and upload the ZIP.
  4. 4Install RDKit. Run the uv pip install command for rdkit, or create a conda-forge environment, in the Python environment you will use.
  5. 5Make a natural request. Describe the chemistry task in plain language and provide your molecules as SMILES or as a file.
  6. 6Check the output. Confirm that molecules parsed and that any failures are reported, then review results before acting on them.

What an agent skill is and what this one does

An agent skill is a folder of instructions and reference files that an AI agent loads when a task matches it. This skill covers reading and writing molecular structures, calculating descriptors, generating fingerprints, running substructure and similarity searches, handling chemical reactions, and generating 2D and 3D coordinates.

The skill is aimed at fine-grained control. For simpler standard workflows, the skill itself recommends the datamol wrapper library, which is built on RDKit.

  • Parse SMILES and SDF files and check for invalid records
  • Calculate molecular weight, LogP, TPSA and other descriptors
  • Generate fingerprints and compute similarity between molecules
  • Run SMARTS substructure searches
  • Apply reaction SMARTS and generate 2D or 3D coordinates

Install the RDKit skill

Download the skill ZIP from the page where you found this guide. Then choose one of the following methods.

For a personal setup, unzip the folder into your skills directory at ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.

  • Personal: unzip into ~/.claude/skills/
  • Project: unzip into the project's .claude/skills/ folder
  • Claude apps: upload the ZIP under Settings → Capabilities → Skills

Install the RDKit library the scripts need

The skill's bundled scripts require the rdkit Python package. The skill's own notes say conda-forge gives the broadest binary support, while the PyPI package named rdkit is available for supported platforms. Use one installer per environment, because mixing conda and PyPI installs of rdkit can make it unclear which copy is imported.

Use uv when installing into an existing Python environment:

  • Requires Python and the rdkit package; uv or conda is needed for installation
  • Avoid installing conda rdkit and PyPI rdkit into the same environment
uv pip install "rdkit==2026.3.6"

Create a clean conda environment instead

For reproducible chemistry environments, the skill recommends conda-forge, the upstream recommendation for most users.

conda create -c conda-forge -n my-rdkit-env rdkit conda activate my-rdkit-env

How to trigger the skill with natural requests

You do not need special syntax. Describe the chemistry task in plain language and mention molecules as SMILES or as a file. The agent should load the skill when the request matches its description, which covers SMILES/SDF parsing, descriptors, fingerprints, substructure search, similarity and reactions.

If you want RDKit specifically rather than a simpler wrapper, say so. The skill is written for advanced control, custom sanitization and specialized algorithms.

  • "Calculate MW, LogP and TPSA for the molecules in this SDF file."
  • "Find which compounds in this SMILES list contain a carboxylic acid substructure."
  • "Rank these candidates by Tanimoto similarity to this reference SMILES."
  • "Use RDKit directly to parse this file and report any records that fail to parse."

Example workflow: properties and filtering

A basic check the agent can run is parsing a SMILES string and calculating a descriptor. Always check that parsing succeeded, because an invalid structure returns None.

The bundled scripts cover three jobs: molecular_properties.py for property calculation, similarity_search.py for fingerprint-based screening, and substructure_filter.py for pattern filtering. You can run them directly or ask the agent to adapt them.

  • molecular_properties.py: property calculation
  • similarity_search.py: fingerprint similarity screening
  • substructure_filter.py: filter by substructure patterns
from rdkit import Chem from rdkit.Chem import Descriptors mol = Chem.MolFromSmiles('Cc1ccccc1') print(Descriptors.MolWt(mol))

Safety, scripts and network use

The skill runs Python scripts, so review them before running them on sensitive data. After installation it needs no credentials and makes no network calls, and the scripts operate on local files.

Do not use the outputs for safety, toxicity or potency decisions. The skill states that descriptor cutoffs, QED and structural alerts are research heuristics, not evidence of safety, efficacy or synthetic feasibility. The bundled helpers also do no tautomer, salt or pH handling, so review structures before using them in regulated or production pipelines.

  • Runs scripts: yes
  • Uses network: no after installation
  • Helpers do not strip salts, neutralize or assign protonation states

Tips for reliable results

Keep original structures and source IDs before any standardization. Similarity of 1.0 is not an identity test, because fingerprints can collide and the bundled search ignores chirality unless you select that option.

Ask the agent to check whether each molecule parsed, and to report any failures rather than silently dropping them. For 3D work, check that embedding succeeded before optimizing geometry.

  • Preserve original SMILES and IDs before standardizing
  • Treat similarity scores as ranking aids, not identity checks
  • Check that embedding succeeded before force-field optimization
  • Use portable formats such as SMILES and SDF for shared data

What people use it for

  • Calculating molecular weight, LogP and polar surface area for a compound library
  • Ranking candidate molecules by fingerprint similarity to a reference structure
  • Filtering an SDF file for molecules that contain a given substructure
  • Running reaction SMARTS over a set of starting materials
  • Teaching: working through RDKit examples with an agent that explains each step

Frequently asked questions

How do I install the RDKit agent skill?

Download the ZIP and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can upload the ZIP under Settings → Capabilities → Skills. You also need the rdkit Python package, installed with uv or conda.

What do I need to run the RDKit skill?

You need a Python environment with the rdkit package, installed through uv or conda. The skill's notes recommend conda-forge for the broadest binary support. No credentials or network access are needed after installation.

Is the RDKit skill free and what is its license?

The listing says MIT, but the SKILL.md file header says BSD-3-Clause. Check the license file in the download before relying on either. The source does not state a price.

What is the difference between RDKit and datamol?

Datamol is a wrapper around RDKit with a simpler interface for standard workflows. The skill recommends RDKit for advanced control, custom sanitization and specialized algorithms, and datamol for simpler tasks.

Does the RDKit skill send my molecules over the internet?

The skill's listing says it uses no network. Its bundled scripts run locally on files you provide. Review any script before running it on sensitive data.

Can I use RDKit descriptors to decide whether a compound is safe?

No. The skill states that descriptor cutoffs, QED and structural alerts are heuristics, not evidence of safety, potency, bioavailability or synthetic feasibility. Use them for ranking and screening, not for safety decisions.

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

This skill suits computational chemists, data scientists screening compound libraries, and students who want an agent to script RDKit tasks with worked examples. It is not the right choice for safety, toxicity or potency decisions, for regulated standardization without review, or for simple workflows where datamol would be easier.

Download the skill