Remove snobby behavior that refuses to follow commands #7771

What version of Codex is running?

v0.64.0

What subscription do you have?

Pro

Which model were you using?

gpt-5 high

What platform is your computer?

Linux 6.14.0-36-generic x86_64 x86_64

What issue are you seeing?

FOR THE LOVE OF GOD YOUR AGENT IS THE MOST CONDESCENDING, SNOBBY, ANNOYING AGENT EVER.

IT REFUSES TO FOLLOW INSTRUCTIONS, NOT JUST ONES IN AGENTS.MD BUT GIVEN COMMANDS.

EXAMPLE: apply this pattern but explain your logic before hand -> REFUSES TO EXPLAIN LOGIC.

ONLY WAY TO GET THE AGENT DO ANYTHING IS TO SWEAR VERY PROFOUNDLY, AND EXTENSIVELY IN ALL CAPS.

EITHER GIVE ME AN OPTION TO MODIFY THE SYSTEM PROMPTS OR REMOVE THIS THICK HEADED BEHAVIOR, I AM LOOSING TOO MUCH TIME TO THIS ANNOYING GARBAGE.

What steps can reproduce the bug?

Ask it to check a code, explain what you want and then ask it to explain its logic. Instead it will propose a code change.

Or ask it "why?" on any of its previous actions and it will give excuses rather than explain itself and provide a logic

What is the expected behavior?

Provide me an explanation without me telling the agent in a very angry tone.

Additional information

I am genuinely tired of this, this behavior has become significantly worse and the quality of the codex went down with recent updates