ds-analysis-plan

What problem does it solve?

This Skill prevents “late discovery” failures by converting a research question into an explicit, reproducible notebook plan with defined cohorts, windows, metrics, validation checkpoints, and decision-ready outputs before any major querying or transformations begin.

Core Features & Use Cases

• Cohort and metric specification: Defines randomization/analysis units, time windows, exclusions, and a concrete metric hierarchy so downstream work is consistent with the decision being made.
• Validation-driven planning: Includes required diagnostics (e.g., balance/SRM/invariants when applicable) and sets a validation level and verification mode to control how strict the outputs must be.
• Operational observability: Requires runtime visibility via compact stage/timing/row-count checkpoints and explicit progress expectations to avoid silent long-running work.
• Reproducibility by default: Produces a plan that another analyst can rebuild and rerun from scratch, with artifacts such as expected output shapes and a final memo.

Quick Start

Use ds-analysis-plan to create a self-contained notebook execution plan for your experiment question, including cohorts, metric computations, validation checks, and the final decision memo.