Checklist of Common mistakes coding AI Agents make

Question

1. Invented Functions or APIs

pandas.read_excel_fast()

Looks real.

Does not exist.

1. Security Vulnerabilities

AI often ignores security unless explicitly asked.

query = "SELECT * FROM users WHERE name = '" + username + "'"

Very dangerous.

1. Code that looks correct but is wrong

supposed to return even numbers

if num % 2 == 1:

return True

Looks clean.

Completely wrong.

1. Inefficient Code

AI usually optimises for readability, not scalability

So it may generate nested loops, repeated database calls

Works for: 10 records

Breaks for: 10 lakh records

1. Outdated Syntax or Deprecated Methods

Training data may contain older code.

So AI may generate code that:

• gives warnings

• is no longer recommended

1. Inconsistent Codebase Style

Example:

camelCase

snake_case

PascalCase

all in same file.

1. Broken Multi-File Integration

AI may generate files independently that:

• don’t connect properly

• mismatch interfaces

Frontend expects:

{ "username": "Nik" }

Backend returns:

{ "user_name": "Nik" }

What other common mistakes have you experienced?