Data Platform Skills

Primary skills for persistent data stores: SQL analytics, schema modeling, data import, and dashboards. Use these for structured, versioned data workflows.

/api/skills/data-platform

5 Skills

data_analysis

Query and analyze persistent data stores using SQL with DuckDB. Primary choice for structured analytics with versioning.

/api/skills/data-platform/data_analysis
Skill Unlocks 3 tools
Content Type
Static (Markdown)
Widgets
n/a

Unlocks Tools

These tools become available when this skill is activated:

data_get_schemadata_list_tablesexecute_shell

Trigger Keywords

querysqlanalyzereportaggregatestatisticsduckdbselect

Instructions Preview

# Data Store Analysis Skill Use this skill when querying and analyzing data in persistent data stores. The platform uses DuckDB, which supports standard SQL with advanced analytical functions. > **When to use this skill:** > - **Use this skill** when data is in a persistent data store (created via `data_modeling` + `data_import`) > - **Use this skill** for SQL-based analytics, reusable queries, or when you need data versioning > - **Use `analyze_spreadsheet`** instead for one-off analysis of uploaded Excel/CSV files via Vertesia document IDs ## Recommended Approach: Query in Sandbox **Always use native DuckDB in the sandbox** for querying data stores. This provides: - Full DuckDB features (extensions, window functions, CTEs) - No result truncation - Direct DataFrame integration - Better performance (no API round-trips per query) - Multi-step analysis in a single session ### How to Query in Sandbox Use `execute_shell` with the `databases` parameter to sync the database: ``` exec...

data_import

Import data from CSV, JSON, Parquet files or inline data with atomic multi-table support

/api/skills/data-platform/data_import
Skill Unlocks 4 tools
Content Type
Static (Markdown)
Widgets
n/a

Unlocks Tools

These tools become available when this skill is activated:

data_importexecute_shelldata_create_tableswrite_artifact

Trigger Keywords

importcsvexcelparquetjsonload datauploadingestetlpreparecleantransform

Instructions Preview

# Data Import Skill Use this skill when importing data into data stores. Supports atomic multi-table imports with automatic rollback on failure. ## Available Tools After learning this skill, you have access to: - `data_import`: Import data from files or inline arrays ## Supported Formats - **CSV**: Comma-separated values with header row - **JSON**: Array of objects or newline-delimited JSON - **Parquet**: Columnar format for large datasets ## Import Modes ### Append Mode (default) Adds new rows to existing data: ``` { "mode": "append", "tables": { "sales": { "source": "gs://bucket/new_sales.csv", "format": "csv" } } } ``` ### Replace Mode Clears table before importing: ``` { "mode": "replace", "tables": { "sales": { "source": "gs://bucket/full_sales.csv", "format": "csv" } } } ``` ## Data Sources ### From GCS ``` { "tables": { "customers": { "source": "gs://my-bucket/data/customers.csv", "format": "csv" ...

data_migration

Perform schema migrations with data transformations - change column types, split/merge columns, restructure tables

/api/skills/data-platform/data_migration
Skill Unlocks 6 tools
Content Type
Static (Markdown)
Widgets
n/a

Unlocks Tools

These tools become available when this skill is activated:

data_querydata_importdata_alter_tabledata_create_tablesexecute_shellwrite_artifact

Trigger Keywords

migratemigrationtransformschema changecolumn typerestructureconvertcast

Instructions Preview

# Data Migration Skill Use this skill when you need to change schemas in ways that require data transformation. Simple changes (add/drop column) can use `data_alter_table` directly, but complex changes need migration. ## When to Use Migration vs Alter | Change Type | Use | |-------------|-----| | Add new column (nullable) | `data_alter_table` | | Drop column | `data_alter_table` | | Rename column | `data_alter_table` | | Change column type | **Migration** | | Split column into two | **Migration** | | Merge columns | **Migration** | | Add NOT NULL constraint | **Migration** | | Restructure table | **Migration** | ## Migration Workflow ### Step 1: Export Current Data ```python # First, query all data from the table # Use data_query tool to get the data, then save to file ``` Or query directly with the tool: ``` data_query: { "store_id": "...", "sql": "SELECT * FROM customers", "limit": 100000 } ``` ### Step 2: Transform with Python First, write the transformation script us...

data_modeling

Create databases and design data store schemas with tables, columns, relationships, and semantic types

/api/skills/data-platform/data_modeling
Skill Unlocks 4 tools
Content Type
Static (Markdown)
Widgets
n/a

Unlocks Tools

These tools become available when this skill is activated:

data_create_databasedata_create_tablesdata_alter_tabledata_get_schema

Trigger Keywords

schematablecolumndata modeldatabasecreate tablealter tableforeign keyrelationshipcreate database

Instructions Preview

# Data Modeling Skill Use this skill when creating databases or designing data store schemas. The data platform uses DuckDB for analytical queries with schema versioning. ## Available Tools After learning this skill, you have access to: - `data_create_database`: Create a new database for storing analytical data - `data_create_tables`: Create new tables with typed columns - `data_alter_table`: Modify existing tables (add/drop/rename columns) ## Schema Design Best Practices ### Table Naming - Use snake_case for table names: `customer_orders`, `sales_data` - Use plural names for entity tables: `customers`, `products` - Use descriptive names for junction tables: `customer_product_access` ### Column Types Available types: - `STRING`: Text data - `INTEGER`: Whole numbers - `DECIMAL`: Precise decimal numbers (financial data) - `BOOLEAN`: True/false values - `DATE`: Date without time - `TIMESTAMP`: Date with time - `JSON`: Structured JSON data - `BLOB`: Binary data ### Semantic Types A...

data_visualization

Create persistent Vega-Lite dashboards with SQL-backed data from data stores. Dashboards use a single query and combined Vega-Lite spec for cross-panel interactivity.

/api/skills/data-platform/data_visualization
Skill Unlocks 10 tools
Content Type
Static (Markdown)
Widgets
n/a

Unlocks Tools

These tools become available when this skill is activated:

data_querydata_preview_dashboarddata_create_dashboarddata_update_dashboarddata_render_dashboarddata_list_dashboardsdata_list_dashboard_versionsdata_snapshot_dashboarddata_promote_dashboard_versiondata_set_dashboard_versioning

Trigger Keywords

dashboardvegachartvisualizationrenderpngpreviewbar chartline chartscatter plotcross-filter

Instructions Preview

# Data Visualization Skill Create Vega-Lite dashboards to visualize data from persistent data stores. Dashboards support cross-panel interactivity (selections in one panel filter another). > **When to use this skill vs `make_chart` (data-analysis):** > - **Use this skill** when data is in a persistent data store and you want saved, reusable dashboards > - **Use `make_chart`** for quick one-off charts or interactive exploration without persistence > **Vega-Lite Reference:** For comprehensive Vega-Lite documentation (marks, encodings, transforms, scales, etc.), see the `vega_reference` skill. ## Workflow: Test Query -> Preview -> Verify -> Create -> Render **IMPORTANT: Always test your query and preview your dashboard before creating it.** 1. **Test query first**: Use `data_query` to verify your SQL returns the expected data 2. **Preview**: Use `data_preview_dashboard` to generate a preview image 3. **Verify the output**: Check that labels are readable, colors are pleasant, layout...