reproducibility-check

name	reproducibility-check
description	Apply the reproducibility checklist to assess whether experiments can be re-run reliably.

Reproducibility Check Skill

When to Use

• You are reviewing the project for reproducibility (e.g. review-paper command with experiments in scope).
• The user wants to know if main results can be reproduced from the repo.

Instructions

• Use references/checklist.md for the checklist items.
• For each main result or experiment claimed in the paper, run through the checklist: entry point, configs, data, metrics, logging, randomness.
• For each item, record: satisfied (with pointer, e.g. "configs/run_experiment.yaml", "scripts/run_experiment.py") or gap (what is missing).
• Output: A short reproducibility report per result (or one table): checklist item → status + pointer or gap. Suggest concrete fixes for gaps (e.g. "add seed to config X").