How can we observe the interaction and thought process of AI CLI

[1129489672-sketch]

How can I observe the interaction and thought process of AI CLI? I want to know the progress.
I utilized it to finalize a proposal for OpenSpec, and the end result was exceptional. However, without an interactive process, I would have been unaware of the progress made or how much longer I had to wait.

[renderful]

Yeah this project seems to have been abandoned like so many others where code is cheap. I'm still using it but will obviously need to maintain my fork. Right now, you can watch the SPEC markdown file you specified. Tasks will be checked off one by one. We need something where we can see the agent output in real time.

[renderful]

In this issue I posted, someone was working on it. It shouldn't be too hard to implement honestly. #39

[1129489672-sketch]

Yes, I have instructed the Cursor to format and output the logs of Cursor. ^_^

[1129489672-sketch]

I've encountered a problem. If it's a large task, and only a portion is completed, it's considered finished, even though it's not actually complete.
For example, a new task.md file is created with only one task: execute an OpenSpec solution, explicitly requiring the complete completion of tasks.md. but only a portion is actually completed.