The summarized message shows only the diff tool view and end turns.
What version of Codex CLI is running?
codex-cli 0.124.0
What subscription do you have?
Free
Which model were you using?
gpt-5.3-codex
What platform is your computer?
Darwin 25.4.0 arm64 arm
What terminal emulator and version are you using (if applicable)?
Ghostty
What issue are you seeing?
The summarized message shows only the diff tool view and end turns. What I find strange is that the diff tool is showing as just colored ANSI text (without a full diff background like edit-diff). The final message showed “Work for N duration” but then showed the diff and ended without any explanation. I thought my token usage was run out, but it's not.
<img width="758" height="905" alt="Image" src="https://github.com/user-attachments/assets/da6b4858-a798-4d77-b458-5929d107e2a3" />
<img width="747" height="706" alt="Image" src="https://github.com/user-attachments/assets/9aa7869b-1b41-4cf6-b6b0-e575832007df" />
<img width="760" height="831" alt="Image" src="https://github.com/user-attachments/assets/499a9575-41dc-46c5-b418-a5b2e556a113" />
What steps can reproduce the bug?
Uploaded thread: 019dc009-353b-7090-9e0c-10e36dd661cb
What is the expected behavior?
Codex should end the turn with a concrete, summarized message, and if the usage quota is run out, it should show a TUI message as normal.
Additional information
_No response_
4 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
<img width="739" height="749" alt="Image" src="https://github.com/user-attachments/assets/c2d3fd5f-03cd-40d3-bdfa-8fb0e9658749" />
New session thread ID 019dc009-353b-7090-9e0c-10e36dd661cb
I just realized another possible cause is that, because /diff is triggered without compact/condensed output, hence, in large diff sessions, the context window could be being overloaded. Hence, it could be that Codex got confused and could not trigger compaction because this is the tool's output, not conversation context => agent's run loop is confused.
Hi @fcoury-oai, any updates on this? :)