---

name: dg:prototype description: Build production-ready Dagster implementations with best practices, testing, and validation. Use when user wants to prototype, build, implement, or create Dagster assets, pipelines, workflows, or data integrations.

Prototype Dagster Implementation Skill

This skill helps users build production-ready Dagster implementations through natural language requests.

When to Use This Skill

Auto-invoke when users say:

• "prototype a dagster pipeline"
• "build a dagster workflow"
• "implement a dagster asset"
• "create a data pipeline in dagster"
• "I want to build [X] using dagster"
• "help me implement [Y] in dagster"
• "create a dagster implementation for [Z]"
• "build assets for [X]"
• "implement [X] with dagster"

How It Works

When this skill is invoked:

1. Extract requirements from the user's request (what they want to build)
2. Identify any specific integrations mentioned (e.g., Snowflake, dbt, APIs)
3. Invoke the underlying command: /dg:prototype <requirements>
4. Handle vague requests:
   • If requirements are too vague, ask clarifying questions
   • Extract data sources, transformations, destinations, and constraints

Example Flows

Clear requirements:

CopyUser: "I want to build a data pipeline that loads data from Snowflake and transforms it with dbt"
→ Invoke: /dg:prototype Build a pipeline that: 1) loads data from Snowflake using dagster-snowflake, 2) transforms with dbt using dagster-dbt

General request:

CopyUser: "Prototype a dagster workflow for processing customer data"
→ Invoke: /dg:prototype Process customer data with appropriate transformations and validation

Asset creation:

CopyUser: "Create a dagster asset that fetches data from an API"
→ Invoke: /dg:prototype Create an asset that fetches data from an external API with error handling and retries

Complex integration:

CopyUser: "Build a pipeline that ingests from PostgreSQL, transforms with Spark, and loads to BigQuery"
→ Invoke: /dg:prototype Build a pipeline that: 1) ingests from PostgreSQL, 2) transforms with PySpark, 3) loads to BigQuery

Vague request - needs clarification:

CopyUser: "Help me build a data pipeline"
→ Ask: "What would you like your pipeline to do? Please describe:
  - Data source(s) (APIs, databases, files, etc.)
  - Transformations needed
  - Output destination(s)
  - Any specific integrations or constraints"
→ Wait for user response
→ Invoke: /dg:prototype <detailed-requirements>

Implementation Notes

• This skill is a thin wrapper that delegates to the /dg:prototype command
• The command file at commands/prototype.md contains the actual execution logic
• The prototype command automatically leverages:
  • dagster-conventions skill for best practices
  • dagster-integrations skill for discovering appropriate integrations
• Explicit slash command invocation still works: /dg:prototype <requirements>
• Both natural language and explicit invocation are supported

Integration with Other Skills

The /dg:prototype command works seamlessly with:

dagster-conventions:

• Provides best practices for asset design
• Guides testing patterns
• Ensures proper resource configuration
• Recommends project structure

dagster-integrations:

• Suggests appropriate integrations for use cases
• Provides integration-specific guidance
• Helps choose between similar tools (e.g., dbt vs Sling)

You don't need to manually invoke these skills - the prototype command handles it automatically.

What Gets Built

When you invoke this skill, the prototype command will:

1. Analyze requirements and suggest appropriate patterns
2. Create production-ready assets with proper dependencies
3. Set up resources and configuration
4. Include comprehensive tests
5. Add documentation and metadata
6. Follow Dagster best practices throughout

Related Commands

Before prototyping, users may need to:

• Use /dg:create-project <name> to set up a new project first
• Navigate to their existing Dagster project directory

After prototyping, users can:

• Run dg dev to see their implementation in the Dagster UI
• Use pytest to run the generated tests
• Extend the implementation with additional assets