The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of CLI tools and Python scripting via the lamin and ln APIs. This includes initialization (lamin init), authentication (lamin login), and instance management (lamin connect), which are necessary for the tool's operation.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the lamindb library and its plugins (e.g., bionty, lamindb-wetlab) from standard registries like PyPI. It also includes instructions for fetching biological ontologies and datasets from established public repositories and cloud storage providers (AWS S3, Google Cloud Storage).
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface that processes external biological datasets (AnnData, Parquet, CSV) and metadata records.
Ingestion points: Files are ingested via methods such as ad.read_h5ad(), pd.read_csv(), and artifact.load() in SKILL.md and reference files.
Boundary markers: The skill does not explicitly define prompt boundary markers for data content, though it uses typed features to scope metadata.
Capability inventory: The skill has access to Bash, Write, and Edit tools, allowing for file modification and command execution based on processed data.
Sanitization: Curation workflows include standardization via cat.standardize() and validation against biological ontologies, which provides a layer of data normalization for biological metadata fields.
[PRIVILEGE_ESCALATION]: Documentation in references/setup-deployment.md includes sudo commands for configuring system-wide cache directories and permissions in multi-user environments. These instructions are transparently documented as administrative setup tasks for shared infrastructure.

alterlab-lamindb