cumcm-data-code-pipeline

name	cumcm-data-code-pipeline
description	Organize reproducible data processing, modeling code, generated outputs, and supporting materials for 2026 CUMCM. Use when creating or reviewing code folders, data provenance, preprocessing scripts, experiment logs, generated figures/tables, appendix code listings, or reproducibility instructions.

CUMCM Data Code Pipeline

Use this skill to make contest code reproducible and submission-safe.

Workflow

1. Separate raw data, processed data, scripts, outputs, and paper assets.
2. Keep generated figures and tables traceable to scripts.
3. Prefer deterministic scripts over notebook-only workflows for final artifacts.
4. Add a support-material file list in the appendix or submission notes.
5. Check for identity leaks and absolute local paths before packaging.
6. Read references/checklist.md for the expected file layout.
7. Read references/directory-layout.md, references/reproducibility-rules.md, and references/support-material-list.md when preparing a real project package.
8. Read references/examples.md when organizing or reviewing an actual code/data folder.

Guardrails

• Do not include private credentials, local usernames, school names, or personal metadata.
• Do not overwrite raw data.
• Do not fabricate data to fill missing contest inputs.