Describe the bug
I scheduled a one-time prompt with something like /after 10m start evals every 100 steps for XYZ train job. Copilot (in my case GPT-5.4 long context medium thinking) executed the task, but the train job wasn't ready enough to launch evals, so it deferred launching until the next schedule tick. However, since /after is a one-time prompt, there won't be another schedule tick; the action I wanted didn't execute. Unfortunately, I can't share the exact prompt or repo I used since they're internal.
Affected version
GitHub Copilot CLI 1.0.61
Steps to reproduce the behavior
- Schedule a prompt with
/after for a task that won't be ready to act on when the task is executed.
- Copilot decides to act on the task "in the next tick" but not execute the action.
Expected behavior
Copilot schedules another later tick to fulfill its promise, or monitors until the right time to execute the action.
Additional context
Windows 11 x86_64, zsh inside zellij through WSL2 on windows terminal. GPT-5.4 medium thinking long context.
Describe the bug
I scheduled a one-time prompt with something like
/after 10m start evals every 100 steps for XYZ train job. Copilot (in my case GPT-5.4 long context medium thinking) executed the task, but the train job wasn't ready enough to launch evals, so it deferred launching until the next schedule tick. However, since/afteris a one-time prompt, there won't be another schedule tick; the action I wanted didn't execute. Unfortunately, I can't share the exact prompt or repo I used since they're internal.Affected version
GitHub Copilot CLI 1.0.61
Steps to reproduce the behavior
/afterfor a task that won't be ready to act on when the task is executed.Expected behavior
Copilot schedules another later tick to fulfill its promise, or monitors until the right time to execute the action.
Additional context
Windows 11 x86_64, zsh inside zellij through WSL2 on windows terminal. GPT-5.4 medium thinking long context.