Tendency to stop the agent on output while it has a plan to move on

What version of Codex is running?

1.105.0 03c265b1adee71ac88f833e065f7bb956b60550a arm64

What subscription do you have?

Pro

Which model were you using?

gpt-5-codex (medium)

What platform is your computer?

macOS

What issue are you seeing?

There's a tendency in the agent to stop at certain points while it basically "wants" to continue (or even thinks it is). There might be some kind of problem to keep the loop going.

I have seen this at several points in projects during a conversational flow but i can also kind of provoke it with a prompt/command i'm working on right now.

Here's some output:

• Updated Plan
└ ✔ Task 1 – Determine whether .ai/constitution.md exists and set project status
✔ Task 2 – Handle optional argument block and record special instructions
□ Task 3 – Scan repository for project essentials
□ Task 4/5 – Choose update or create flow actions
□ Task 6 – Validate planned constitution output
□ Task 7 – Save constitution and report final facts

─ Worked for 27s

• Task 1 ✅ Status: FRESH — .ai/constitution.md is missing, so we’ll draft a new
constitution. 😊
Task 2 ✅ Status: No special instructions were supplied. Moving on to the project scan
next.

---

That's where it ended.

Both Tasks at the end had this instruction in the prompt:

- **PRINT STATUS AND CONTINUE**

It works about 30-50% of the times with instructions like this. If I ask backed why it stopped it said it didn't and it's already continuing with the next step just to stop again.

Not sure if the model causes this but i would assume it's probably more on the codex side.

What steps can reproduce the bug?

Maybe try a prompt file, give it structured tasks and clear instructions to follow through without pause - doesn't matter. I don't think the instruction really changes this.

What is the expected behavior?

Continuing following the prompt without stopping.

Additional information

_No response_