BERIL-research-observatory

Run an independent AI review of a project or research plan. Use when you want feedback without the full /submit checklist.

Approve a project and upload it to the lakehouse. Use when the author is ready to stand behind the report and submit the project for archival.

Read analysis outputs, compare against literature, and draft findings for a project REPORT.md. Use when notebooks have been run and the user wants to interpret results and write up findings.

Get started with the BERIL Research Observatory. Use when a user is new, wants orientation, or asks what they can do.

Run arbitrary scripts on KBase compute nodes via the CDM Task Service (CTS). Use when the user needs to move compute off their notebook or local machine — e.g., running bioinformatics tools, heavy data processing, or anything that benefits from dedicated CPU/memory on a remote node.

Use when searching BERIL project/docs context through OpenViking or refreshing the indexed context layer before research, synthesis, or pitfall work.

Search and review biological literature using MCP tools (PubMed, arXiv, bioRxiv, Google Scholar) with full-text reading, citation snowballing, and PaperBLAST integration. Use when the user wants to find papers, review existing research on a topic, check what's known about an organism or pathway, or support a hypothesis with citations.

Structural biology workflow orchestrator using the Phenix suite. Use when the user wants to determine, refine, or validate protein structures — including AlphaFold structure retrieval, X-ray crystallography, cryo-EM, MolProbity validation, or visualization script generation.

Ingest a dataset into the BERDL Lakehouse from within JupyterHub (in-cluster). Data may live on the JH filesystem or a global shared filesystem. Handles schema detection, MinIO upload via Python client, and Iceberg table creation via the data_lakehouse_ingest pipeline. Use when a user is already working inside JupyterHub and wants to load a new dataset — SQLite, TSV, CSV, Parquet, or other tabular formats — into a Lakehouse namespace. For off-cluster ingestion from a local machine, use berdl-ingest-remote instead.

Ingest a local dataset into the BERDL Lakehouse from a local (off-cluster) machine via SSH tunnels and pproxy. Handles data format detection and preparation, MinIO upload, and Iceberg table creation via the data_lakehouse_ingest pipeline. Use when a user wants to load a new dataset — SQLite, TSV, CSV, Parquet, or other tabular formats — into a Lakehouse namespace from their local machine (not from within JupyterHub). For in-cluster ingestion from within JupyterHub, use berdl-ingest instead.

Detect and document pitfalls encountered during BERDL work. Invoked by other BERDL skills when errors, retries, or data surprises occur.

Review completed projects and their findings, then suggest a new high-impact research topic grounded in available BERDL data and scientific gaps. Use when the user wants to identify the next best research direction based on what has already been done.

Run SQL queries from a local machine against a provisioned BERDL Spark cluster using spark_connect_remote. Use when the user wants remote Spark compute with local control, needs clarity on connection and timeout behavior, or wants to return small/medium results directly before exporting large outputs.

Query the KBase BERDL (BER Data Lakehouse) databases. Use when the user asks to explore pangenome data, query species information, get genome statistics, analyze gene clusters, access functional annotations, or query biochemistry data.

Discover and document BERDL databases. Use when the user wants to explore a new database, generate documentation for a database, or create a module file for the berdl skill.

Retrieve and use BERDL MinIO credentials and transfer result artifacts between BERDL object storage and the local machine. Use when exported query results need to be listed, downloaded, shared, or when only KBASE_AUTH_TOKEN is available and MinIO keys must be acquired.

Generate LinkML schema YAML from markdown, Excel, or text descriptions. Scaffold a LinkML project repo and push to GitHub.