If you need to manipulate molecular structures at a granular level, this is the canonical cheminformatics library. It handles SMILES and SDF parsing, calculates descriptors like molecular weight and LogP, generates fingerprints for similarity searches, and manages 2D/3D coordinate generation. The skill documentation covers sanitization control (RDKit runs 13 validation steps by default), stereochemistry assignment, and the modern fingerprint generator API for Morgan and RDKit topological fingerprints. Worth noting that the maintainers recommend using datamol for standard workflows since it wraps RDKit with a simpler interface. Reach for this when you need custom sanitization logic, specialized substructure algorithms, or direct control over molecular validation steps that higher-level tools abstract away.
npx skills add https://github.com/k-dense-ai/scientific-agent-skills --skill rdkit