Open Record Step 10: Clean It, In a Script

Write clean.py: read from data/raw, write to data/clean, never modify the raw files. Normalize the categorical columns, convert the numeric ones, and handle missing values with a decision recorded per column in a comment. Print a summary of every change: rows in, rows out, values normalized, rows dropped and why. Then read the summary and ask whether any single decision is doing more work than you are comfortable with. If dropping missing rows removed a tenth of your data, that is a finding, not a cleanup step.