The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the cellxgene-census package from public registries and includes code examples that fetch standardized genomic data from the CZ CELLxGENE Census remote repository for analysis.
[COMMAND_EXECUTION]: The skill documents the use of uv pip for installing the necessary Python dependencies required to interact with the Census API and machine learning frameworks.
[DATA_EXFILTRATION]: While the skill involves reading large volumes of genomic data, it does not contain any patterns suggesting data exfiltration. The data is processed locally using standard libraries such as scanpy and PyTorch.
[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing external data from the Census repository.
Ingestion points: Data is ingested via cellxgene_census.get_obs(), cellxgene_census.get_var(), and cellxgene_census.get_anndata() in SKILL.md.
Boundary markers: None present; the skill assumes the data follows the SOMA/AnnData structured schema.
Capability inventory: The skill uses data science libraries (pandas, numpy, scanpy, torch) to manipulate the ingested data.
Sanitization: The skill relies on the structured nature of genomic data and the cellxgene-census API for data integrity. The risk is considered minimal due to the scientific nature of the data source.

cellxgene-census