Nrakow Ae Skills Dev New Project Setup

When to Use This Workflow

Use new-project-setup when starting an analytics engineering project from zero: no dbt project exists yet, you're onboarding a new team, or you're rebuilding a project from scratch. If a dbt project already exists, use individual skills instead of this workflow.

Overview

This workflow sequences six skills in order. Each phase builds on the previous one. Complete each phase before moving to the next.

Phase 1: Capture Stack Context

Skill: data-stack-context

Before writing any code, capture the team's full analytics stack. This file drives all subsequent skill decisions.

What to do:

1. Check if .claude/data-stack-context.md already exists. If it does, skip to Phase 2.
2. Invoke the data-stack-context skill.
3. Use Path A (auto-draft) if dbt_project.yml or profiles.yml already exist. Use Path B (conversational) if starting from zero.
4. Confirm the context file is written to .claude/data-stack-context.md.

Phase complete when: .claude/data-stack-context.md exists and covers warehouse, dbt version, orchestrator, and BI layer.

Phase 2: Scaffold the dbt Project

Skill: dbt-project-setup

Set up the project structure, profiles, packages, and schema naming conventions.

What to do:

1. Invoke the dbt-project-setup skill.
2. Generate dbt_project.yml, profiles.yml, packages.yml, and the generate_schema_name macro.
3. Run dbt deps && dbt debug to confirm connectivity.

Phase complete when: dbt debug passes and dbt parse returns no errors.

Phase 3: Establish SQL Style Guide

Skill: sql-style-guide

Lock in formatting and naming standards before writing any models — it's much harder to enforce retroactively.

What to do:

1. Invoke the sql-style-guide skill.
2. Generate a .sqlfluff config tailored to the team's warehouse dialect.
3. Document key conventions (CTE naming, column aliasing, join style).

Phase complete when: .sqlfluff config exists and sqlfluff lint runs without configuration errors.

Phase 4: Design the Data Model Architecture

Skill: data-modeling

Plan the layer structure and identify initial domains before writing the first model.

What to do:

1. Invoke the data-modeling skill.
2. Identify 2-3 initial data domains (e.g., customers, orders, events).
3. Sketch the staging → intermediate → mart flow for each domain.
4. Run node tools/clis/schema-introspect.js if source schemas are already accessible.

Phase complete when: An entity-relationship sketch and initial model list exist for the first domain.

Phase 5: Set Up Testing Strategy

Skill: data-quality-testing

Establish the testing standard before writing models — define what "good" means.

What to do:

1. Invoke the data-quality-testing skill.
2. Define the minimum test set per model tier (staging, intermediate, marts).
3. Add Elementary or Soda if the stack includes an observability tool.

Phase complete when: Testing standards are documented and at least one example schema.yml with tests is written.

Phase 6: Configure CI/CD

Skill: dbt-ci-cd

Automate testing and deployment so the project is production-ready from day one.

What to do:

1. Invoke the dbt-ci-cd skill.
2. Set up a CI workflow that runs dbt build --select state:modified+ on PRs.
3. Configure production job scheduling in dbt Cloud or the orchestrator.

Phase complete when: A CI pipeline exists and a test PR triggers a dbt build automatically.

Final Verification

dbt debug        # Confirm warehouse connectivity
dbt parse        # Confirm no compilation errors
dbt deps         # Confirm packages installed
sqlfluff lint models/   # Confirm style rules enforced

Verify Your Work

Do not present output from this skill as complete until every command below passes without error. If a command fails, consult "If Something Goes Wrong" before asking the user.

• Run dbt debug to confirm warehouse connectivity and that profiles.yml is correctly configured for your target environment.
• Run dbt parse to confirm no compilation errors exist across all model paths defined in dbt_project.yml.
• Run dbt deps to confirm all packages in packages.yml resolved and installed without version conflicts.
• Run sqlfluff lint models/ to confirm the .sqlfluff config is valid and style rules are enforced without configuration errors.
• Open a test PR and confirm the CI pipeline triggers automatically and runs dbt build --select state:modified+.

If Something Goes Wrong

• dbt debug fails: check profiles.yml path and warehouse credentials. Ensure env vars are set for the correct target.
• dbt parse fails: usually a missing ref() or misconfigured model path in dbt_project.yml. Check model-paths setting.
• dbt deps fails: check packages.yml for version conflicts. Run dbt deps --upgrade to resolve.
• CI never triggers: confirm the workflow file is in .github/workflows/ and the branch protection rule runs the check.

Source: nrakow/ae-skills-dev — distributed by TomeVault.