One core habit
Treat files as the input layer of your tool. Read the file, convert it into Python data, then let the rest of your code work with lists and dictionaries.
Do not mix file reading, API calls, and output writing in one big block. Your future self will find bugs faster when each part has one job.
Why this matters
Your hackathon repo must prove that your script reads data. A sample input file is the easiest proof.
For an AI-adjacent tool, the input may be student feedback, product reviews, job descriptions, support tickets, or prompts. Keep a small sample in the repo so anyone can test it without asking you for data.