OpenAI Documentation Skill GET RID OF! MASSIVE TOKEN BURN!!!

What version of the Codex App are you using (From “About Codex” dialog)?

Version 26.818.61809

What subscription do you have?

200 pro

What platform is your computer?

Microsoft Windows NT 10.0.26200.0 x64

What issue are you seeing?

A Codex System Skill for OpenAI Documentation has been added apparently and is a MASSIVE TOKEN BURN. I have sent ticket in the past regarding the absolutely inefficient website structure for Agent documentation lookup and retrieval on the OpenAI website documentation pages. I have had to do workarounds to create markdown versions of these pages so my agents would stop using the website because it would burn through 40% of my context window for that agent in a couple minutes if i'm not looking at it. For the smallest and simplest questions if i ask anything about Codex. And now that it's a system skill, it's overriding my instructions for my agents and they now go straight to the website again. That skill needs to be removed until your website or a retrieval method or the protocol your skill does in it is better designed.

What steps can reproduce the bug?

Ask an agent anything about Codex and it will run the skill and be directed to the website for lookups. Then it burns through your tokens faster than you can blink.

What is the expected behavior?

It shouldn't use up tokens like that for it's own documentation.

Additional information

FIX THIS!

Reproduce the issue by asking an agent a Codex question and observe whether the OpenAI Documentation system skill is invoked and consumes excessive context. Done means the skill no longer causes disproportionate token use for simple Codex documentation questions, while the expected documentation lookup behavior remains available.