Default command to run for new terminal threads

[PrayagS]

What are you proposing?

Originally discussed here: https://discord.com/channels/869392257814519848/1504795790281932890

For every new terminal agent thread, there should be a default command that's executed. Eg. claude. This prevents the user from having to type that out for every thread.

This can be modeled as different profiles of terminal agent threads so user can spawn different harnesses.

[macraig]

This is a really cool idea! I'll share it with the team to see what we can do

[nickbreaton]

I've been using the terminal agent a lot more across projects, and not having something like this is creating a lot of friction. Closing and reopening Zed can cause all my terminal sessions to reset back to the shell.

Being able to run something like pi --continue (or similar in your agent harness of choice) via a hook would allow things to resume much easier when jumping back in.

[nickbreaton]

If this is implemented as hooks, it could be nice to even provide more of a granular hook for the resume state (i.e. terminal_thread_create, terminal_agent_restore). Pi handles this gracefully with its continue flag, restoring if possible, but I'd assume some other agent harnesses many not.

[macraig]

Converted this to an issue that we're tracking here: #58697

[macraig]

Shipped in #59374