Replication And Research Hygiene

Trigger

Use when the task involves:

• making results rerunnable by someone else
• deciding which scripts or notebooks are canonical
• packaging tables, figures, simulations, or model outputs for handoff
• checking for hidden path, seed, dependency, or environment assumptions

Workflow

1. Inventory inputs, outputs, environment assumptions, and any required secrets or credentials.
2. Distinguish exploratory artifacts from canonical pipelines.
3. Verify paths, seeds, configuration values, and result-generation steps.
4. Write a short replication note with commands, dependencies, expected outputs, and remaining blockers.

Guardrails

• Do not claim a result is reproducible if it depends on undocumented manual edits.
• Avoid absolute paths and hidden local dependencies.
• Keep raw data immutable.
• Name the canonical script or entry point that regenerates each table, figure, or model output.