starlake-skills

Manage GizmoSQL processes: start, stop, list, and stop-all DuckLake-backed SQL servers

Create or modify database connections in application.sl.yml

Manage Quack DuckDB query servers exposing DuckLake over a thin remote protocol — serve (foreground), start/stop/list/stop-all (background)

Automatically infer schemas and load data from the incoming directory

Data quality expectations syntax, built-in macros, and validation patterns

Apply Row Level Security (RLS) and Column Level Security (CLS) policies

Run SQL or Python transformation tasks

Generate SQL DDL statements from Starlake YAML definitions

Adversarial parallel review of pipeline changes. Spawns Winston (architect), Amelia (engineer), and Quinn (data quality) as independent subagents, then triages findings. Use when the user says "review pipeline", "code review", or "review this data change".

Design the overall data architecture including layers, storage, engines, and governance. Use when the user says "create data architecture" or "design the data platform".

Create a complete pipeline specification covering extract, load, transform, and orchestrate. Step-file workflow with resumable progress. Use when the user says "create pipeline spec" or "design a data pipeline".

Business Data Analyst agent: guides domain discovery and source analysis. Use when the user says "data-analyst" or "talk to the data-analyst".

Data Architect agent: designs data platforms, schemas, and pipeline architecture. Use when the user says "data-architect" or "talk to the data-architect".

Data Engineer agent: builds and maintains ETL/ELT pipelines with Starlake. Use when the user says "data-engineer" or "talk to the data-engineer".

Data Quality Engineer agent: ensures data integrity with expectations, lineage, and governance. Use when the user says "data-quality-engineer" or "talk to the data-quality-engineer".

Implement a data pipeline from a pipeline specification, generating Starlake configuration files. Use when the user says "implement pipeline" or "dev this pipeline".

Discover and document data domains, sources, and ownership. Use when the user says "discover data domains" or "map data sources".

Reads the Starflow manifest and scans existing artifacts to tell the user where they are and what to do next. Use when the user asks "what should I do next", "starflow help", "where am I in starflow", or just "/help" inside Starflow.

Review and document data lineage across pipeline stages. Use when the user says "review lineage" or "trace data flow".

Design orchestration DAGs for scheduling and managing data pipeline execution. Use when the user says "design orchestration" or "create DAG configuration".

Platform Engineer agent: manages infrastructure, orchestration, and deployment for data pipelines. Use when the user says "platform-engineer" or "talk to the platform-engineer".

End-of-epic retrospective for a data pipeline batch. Pulls the previous retro and checks follow-through on action items, then captures wins/struggles/patterns and produces new action items with clear ownership. Use when the user says "run a retro", "epic retrospective", or "retro the {epic}".

Design Starlake-compatible table schemas with types, constraints, privacy, and expectations. Use when the user says "design schema" or "create table definition".

Analyze data sources in depth: schema, quality, volume, and extraction strategy. Use when the user says "analyze data source" or "profile this data source".

Design SQL transformations for data pipelines with quality checks and dependency management. Use when the user says "design transforms" or "create SQL transformations".

Extract data from database tables to CSV/Parquet files

Extract data from REST API endpoints into CSV or JSON Lines files with pagination, auth, incremental, resume, proxy, and mTLS support

Extract schemas from REST API endpoints by fetching sample responses and inferring data structure

Extract both schema and data from a JDBC source

Generate ACL (Access Control List) dependencies graph

Create a new Starlake project from a template

Get table information from BigQuery

Load files (Parquet/CSV/JSON) into a JDBC table

Generate column-level lineage for a specific task

Compare two versions of a Starlake project

Start the Starlake interactive REPL console

Deploy generated DAGs to a target directory

Generate orchestration DAGs (Airflow/Dagster) from your Starlake project

Load data into Elasticsearch

Extract schemas directly from BigQuery datasets