Display authorized commands

What variant of Codex are you using?

VsCode Extension

What feature would you like to see?

Identifier openai.chatgpt
Version 0.4.79

019cb4ad-cb74-77d0-955e-bd7373f1fc34

So, I wasn't sure where or how to post this, but I thought that reporting how it's actually working is actually a good thing as I would like to see it do more of this.

I was finally able to get Codex to eventually run with minimal supervision by getting it to use the same smaller commands which I can approve just once, rather than it use monolithic multiple commands on each command line. This took a long time to get into this groove and want to share it so that you could get it to do this more consistently, because honestly, this was an uphill battle and shouldn't have been this difficult.

I actually have a repo (AI-General-Memory) that is trying to get more consistent and intelligent responses from both Codex and Claude.

I'm not saying that it's not executing a different command now and then, but right now it's doing it far less, allowing a migration that could have taken days to collapse to just 1 or less.

Here is most of the project CODEX.md file that I use to remind Codex.

CODEX.md

Feature Request

On that note, I've noticed that after compaction, the commands that I authorised get kicked, which is annoying as I have to re-authorised them. It would be also nice feature to be able to see the authorised commands (and maybe the unauthorised ones) so that I could review them and disable them in case I inadvertently allowed it and don't want it to do it again without me reauthorising it again. Or conversely, I disallowed it and didn't mean to.

I hope to see more of this in the future! Cheers!

Additional information

Keep up the good work!!! 😊