The Agent Skills Directory

[SAFE]: The skill provides integration for the LaminDB framework, facilitating biological research through data queryability, lineage tracking, and ontology standardization.
[EXTERNAL_DOWNLOADS]: Instructs the installation of the lamindb package and its ecosystem (e.g., bionty, wandb, anndata) from official registries using standard package managers.
[COMMAND_EXECUTION]: Includes administrative CLI commands for initializing environments and authenticating users (lamin login, lamin init), which are necessary for the framework's operation.
[DATA_EXFILTRATION]: Supports the transfer of research data to well-known cloud storage providers (S3, Google Cloud Storage) and MLOps platforms (Weights & Biases, MLflow) as part of intended scientific workflows.
[REMOTE_CODE_EXECUTION]: References standard patterns for loading serialized data artifacts and machine learning models (e.g., via joblib or artifact.load()). These operations are consistent with scientific data processing but involve inherent deserialization risks associated with the data formats used.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by ingesting external biological datasets. Ingestion points include artifact.load() and ad.read_h5ad. The skill documentation does not define specific boundary markers or content sanitization for the ingested data before it is processed within the agent's context.

lamindb