Agent thinks that a tool call that waited for user confirmation "took longer than expected"

What version of Codex CLI is running?

codex-cli 0.151.0

What subscription do you have?

pro

Which model were you using?

gpt-5.6-sol

What platform is your computer?

linux

What terminal emulator and version are you using (if applicable)?

kitty

Codex doctor report

What issue are you seeing?

If I take a while to approve an out-of-sandbox execution the agent will often say afterwords that "such and such operation is taking longer than expected" or "xyz command took longer to run locally than in CI" or similar. There is something about the tool call where the harness is reporting to the agent a runtime or start/end time that is artificially extended by the time it takes me to respond to the approval prompt.

What steps can reproduce the bug?

There's no guaranteed way but in general if you have the sandbox setup and the agent needs to prompt you all the time to do some actions and you don't respond promptly it might make some comment about the operation taking longer than expected and often hallucinate some reason why that is without any awareness it is because of the wait for confirmation.

What is the expected behavior?

I think if the agent is being given information about the duration of the tool call it should be given accurate information that distinguishes the time spent waiting for user input versus the time the tool actually took to execute.

Additional information

No response

Start by tracing the Codex CLI approval prompt and the tool-call timing reported to the agent. Reproduce the issue by delaying approval for an out-of-sandbox execution, then verify that the reported duration distinguishes user-confirmation time from actual command execution time.