lamindb — Claude Code skill: biological data curation

What lamindb does

The skill gives your agent a working procedure for LaminDB 2.10.0 and Bionty 2.5.0. It defines typed Feature records, builds a Schema from them, runs curator.validate() on a table and saves the result as an artifact. Collections, branches, spaces and run lineage are covered too, so you can see which inputs and outputs a script left behind.

A good part of it is a list of traps. Artifact.backed(), delete_cache() and is_cached() do not exist in this release. An open() call on Parquet hands back a PyArrow dataset rather than a byte stream, and curator.cat.add_ontology() is gone — set the field's Bionty dtype and call cat.standardize() instead. For AnnData the path is curator.slots["obs"].cat.

You need Python 3.10–3.14 and lamindb itself; the examples were run on Python 3.12. A local SQLite instance needs no login. Public ontologies and remote storage need network access, and private instances need credentials. Registering an artifact says nothing about whether the science holds up — assay-specific checks are yours to write.