Claude Skills User Guide

This guide explains how to use Claude Code's Slash Commands to quickly generate and configure TM/QM model files.

What are Claude Skills?

Claude Skills are predefined prompt templates integrated into Claude Code as Slash Commands. When you type /command-name, Claude loads the corresponding specialized knowledge and rules to help you complete specific tasks more efficiently.

This project provides two TM/QM related Skills:

Command	Function
/tm-generate	Generate complete TM files from DDL or table descriptions
/tm-dimension	Configure dimension relationships (basic, nested, parent-child)

Prerequisites

1. Install Claude Code
2. Run Claude Code in the project root directory
3. Ensure the skill files exist in .claude/commands/ directory

/tm-generate - TM File Generation

Overview

Automatically generate TM (Table Model) files based on user input, supporting:

• DDL statement parsing
• Natural language table descriptions
• Database table names (combined with dataset.inspect_table tool)

Usage

In Claude Code, type:

/tm-generate

Please generate a TM file for the following table:

CREATE TABLE fact_sales (
    sales_key BIGINT PRIMARY KEY,
    date_key INT NOT NULL,
    customer_key INT NOT NULL,
    product_key INT NOT NULL,
    order_id VARCHAR(50),
    quantity INT,
    unit_price DECIMAL(10,2),
    sales_amount DECIMAL(12,2)
);

Smart Inference

/tm-generate automatically:

1. Type Mapping: Converts database types to TM types

   • BIGINT → BIGINT
   • DECIMAL → MONEY
   • VARCHAR → STRING
   • DATE → DAY

2. Role Detection:

   • Foreign key columns → dimension
   • Amount/quantity columns → measure
   • Other columns → property

3. Dimension Reuse Suggestions:

   • Identifies common dimensions (date, customer, product)
   • Suggests using dimension builder functions

Output Example

import { buildDateDim, buildCustomerDim, buildProductDim }
    from '../dimensions/common-dims.fsscript';

export const model = {
    name: 'FactSalesModel',
    caption: 'Sales Fact Table',
    tableName: 'fact_sales',
    idColumn: 'sales_key',

    dimensions: [
        buildDateDim({ name: 'salesDate', foreignKey: 'date_key' }),
        buildCustomerDim({ foreignKey: 'customer_key' }),
        buildProductDim({ foreignKey: 'product_key' })
    ],

    properties: [
        { column: 'sales_key', caption: 'Sales Key', type: 'BIGINT' },
        { column: 'order_id', caption: 'Order ID', type: 'STRING' }
    ],

    measures: [
        { column: 'quantity', caption: 'Quantity', type: 'INTEGER', aggregation: 'sum' },
        { column: 'unit_price', caption: 'Unit Price', type: 'MONEY', aggregation: 'avg' },
        { column: 'sales_amount', caption: 'Sales Amount', type: 'MONEY', aggregation: 'sum' }
    ]
};

Combined with Database Inspection Tool

If your database table already exists, combine with dataset.inspect_table:

/tm-generate

Please inspect the fact_sales table in the database and generate a TM file.

Claude will first call dataset.inspect_table to get the table structure, then generate the TM file.

/tm-dimension - Dimension Configuration Helper

Overview

Helps configure various types of dimension relationships:

• Basic Dimension: Direct relationship between fact and dimension tables
• Nested Dimension (Snowflake): Multi-level dimension hierarchy
• Parent-Child Dimension (Hierarchy): Tree structure using closure tables

Usage

/tm-dimension

I need to configure a product dimension where products link to categories,
and categories link to category groups.

Basic Dimension Configuration

{
    name: 'customer',
    caption: 'Customer',
    tableName: 'dim_customer',
    foreignKey: 'customer_key',    // FK in fact table
    primaryKey: 'customer_key',    // PK in dimension table
    captionColumn: 'customer_name',

    properties: [
        { column: 'customer_id', caption: 'Customer ID' },
        { column: 'customer_type', caption: 'Customer Type' }
    ]
}

Nested Dimension Configuration (Snowflake)

IMPORTANT: In nested dimensions, foreignKey refers to the column in the PARENT table, not the fact table!

{
    name: 'product',
    tableName: 'dim_product',
    foreignKey: 'product_key',      // Column in FACT table
    primaryKey: 'product_key',
    captionColumn: 'product_name',

    dimensions: [                    // Nested sub-dimensions
        {
            name: 'category',
            alias: 'productCategory',
            tableName: 'dim_category',
            foreignKey: 'category_key', // Column in dim_product!
            primaryKey: 'category_key',
            captionColumn: 'category_name'
        }
    ]
}

Parent-Child Dimension Configuration (Hierarchy)

Requires a pre-created closure table:

{
    name: 'team',
    tableName: 'dim_team',
    foreignKey: 'team_id',
    primaryKey: 'team_id',
    captionColumn: 'team_name',

    // Parent-child specific configuration
    closureTableName: 'team_closure',  // Required
    parentKey: 'parent_id',
    childKey: 'team_id'
}

Dimension Reuse Patterns

/tm-dimension suggests extracting common dimensions as builder functions:

// dimensions/common-dims.fsscript
export function buildDateDim(options = {}) {
    const { name = 'date', foreignKey = 'date_key' } = options;
    return {
        name,
        tableName: 'dim_date',
        foreignKey,
        primaryKey: 'date_key',
        properties: [
            { column: 'year', caption: 'Year' },
            { column: 'month', caption: 'Month' },
            { column: 'day_of_week', caption: 'Day of Week' }
        ]
    };
}

Best Practices

1. Inspect Before Generate

For existing database tables, inspect the structure first:

First use inspect_table to check the fact_orders table,
then use /tm-generate to create the TM file

2. Prioritize Dimension Reuse

When multiple fact tables use the same dimension, prefer dimension builders:

/tm-dimension

Please create a reusable date dimension builder,
I have multiple fact tables that need it

3. Configure Complex Dimensions Step by Step

For nested or parent-child dimensions, configure in steps:

/tm-dimension

Step 1: Configure the basic product dimension structure
Step 2: Add the category nested dimension
Step 3: Add the category group second-level nesting

4. Provide Business Context

Providing business context results in better field naming and descriptions:

/tm-generate

This is an e-commerce order detail table for sales analysis:
- date_key: order date
- customer_key: ordering customer
- sales_amount: order amount (needs to be aggregated by date and customer)

FAQ

Q: Skill not loading?

Ensure .claude/commands/tm-generate.md and .claude/commands/tm-dimension.md files exist.

Q: How to customize a Skill?

Edit the markdown files in the .claude/commands/ directory directly.

Q: Can I add more Skills?

Yes. Create a new .md file in .claude/commands/ directory. The filename becomes the command name.

Related Documentation

• TM Syntax Reference
• QM Syntax Reference
• Dimension Reuse Best Practices
• Parent-Child Dimension Configuration