Cleaning the Reply

parsing the model's bullet

Part of: Resume Bullet Booster

You asked for a bare bullet. Sometimes you get one. Sometimes you get - Managed the front desk, or "Managed the front desk", or 1. Managed the front desk, or a cheerful Here's your bullet: glued to the front. The model is helpful in ways that break your tool. This lesson builds the cleanup step that turns any of those into one plain bullet. Why parsing is a separate step Never trust the raw reply. Even with a strict prompt, models drift. They wrap text in quotes, add a leading dash because "that's what bullets look like," or number the line. Your downstream code (the action-verb check, the length check) expects a clean string that starts with the actual first word. A stray - makes the "first word" a dash, and every check misfires. So right after the call, you normalize. The order of operations Cleanup is a pipeline, and order matters: 1. Strip whitespace first, so a leading space doesn't hide the marker. 2. Remove one bullet marker (- , , • , or a numeric 1. ) if present. Just one, a real bullet won't start with two. 3. Strip surrounding quotes if both ends have them. 4. Strip again : because removing a marker can expose new whitespace. Each step is dull on its own. Run them in seq

Challenge: Strip the Bullet