ds-planning-workflows

name	ds-planning-workflows
description	Create and manage structured JSON planning files for data science workflows. Use when user requests DS projects requiring data contracts, evaluation protocols, and MLOps requirements.

DS Planning Workflows

When to use this skill

Use this skill when the user requests data science projects that require:

• Data engineering pipelines with quality contracts
• Model development with evaluation protocols
• Production ML deployment with monitoring
• End-to-end DS workflows (DE → DS → MLE)
• Projects with leakage risks or temporal validation needs

Quick start

# Create DS plan
python scripts/create_ds_plan.py --complexity moderate --task "description"

# Validate DS-specific requirements
python scripts/validate_ds_plan.py plan-ds-[name].json
python scripts/check_leakage_risks.py --plan plan-ds-[name].json

Need DS complexity help? → ds-complexity-guidelines.md
Want to see DS examples? → examples/ (real DS project files)

DS Planning templates by complexity

Level	Duration	Template	Usage
Simple DS	2-4 weeks	simple-ds-project.json	Single model, basic validation
Moderate DS	4-12 weeks	moderate-ds-project.json	Data contracts + deployment
Complex DS	12-24 weeks	complex-ds-project.json	Multi-model, production ML
Enterprise DS	24+ weeks	enterprise-ds-project.json	Full MLOps + governance

Unsure which level? → Use DS complexity assessment

Core DS workflow

1. Analyze DS complexity → Select DS template
2. Generate DS plan → Include data contracts + evaluation protocols
3. Validate DS structure → Check leakage risks + reproducibility
4. Execute with DS tracking → Follow DE→DS→MLE order
5. Monitor DS progress → Use DS-specific validation gates

Need DS JSON structure details? → ds-examples.md
Want DS collaboration patterns? → ds-examples.md

DS Agent roles

• @head-of-ds-router: Orchestration with decision rights
• @data-engineer: Data contracts, pipelines, DQ
• @data-scientist: Features, baselines, evaluation
• @ml-engineer: Production, monitoring, deployment
• @ds-validator: Leakage detection, reproducibility

Planning DS agent handoffs? → ds-examples.md

Resources by DS use case

🤔 "How complex is my DS project?" → ds-complexity-guidelines.md
📋 "Show me real DS examples" → examples/ (churn, MMM, recommendation projects)
🔧 "How do DS agent handoffs work?" → ds-examples.md
🏗️ "What's the DS JSON structure?" → ds-examples.md
⚠️ "DS project went wrong, what to avoid?" → ds-examples.md

DS Progress tracking

⚠️ Note: Use ds_progress_tracker.py for DS-specific validation gates

# Track DS completion with validation
python scripts/ds_progress_tracker.py --plan [file].json --complete-step N --validate

# Check leakage risks
python scripts/check_leakage_risks.py --plan [file].json --step N

Creates [plan-name].ds-progress.json alongside your planning file to track DS validation status.

DS-Specific Features

Data Contracts

• Schema definitions with constraints
• Data quality SLAs (>95% pass rate)
• Lineage and refresh requirements
• Interface specifications DE→DS→MLE

Evaluation Protocols

• Baseline establishment requirements
• Leakage prevention validation
• Temporal validation strategies
• Statistical significance testing

MLOps Requirements

• Reproducibility standards (<1% variance)
• Model registry and versioning
• Monitoring and drift detection
• Deployment and rollback strategies

Risk Assessment

• Data leakage risks
• Reproducibility challenges
• Production scaling constraints
• Business impact validation