[BUG] SubAgents Expansions Broken (breaks terminal in Cursor) - with Suggestion on fix

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Hello,

Working with sub-agents in Claude code is handy, but to see what they're doing is a pain, as "ctrl+r" to expand ends up showing a ton of info which causes my terminal to malfunction / flicker / go blank / etc. So I avoid "ctrl+r" as it often makes my terminal unusable and I have to close my Cursor IDE and reopen altogether.

It'd be great if I could have other options from the main agent than just "ctrl+r", that allows me to expand all tool uses done, for example, and from there expand certain ones like all writes, reads, etc. Or to expand all thoughts (with tool uses collapsed or something so I can see the flow of conversation), etc. Just some means to better visualize what my agent's doing without flooding my terminal with a ton of text / info that causes the terminal to malfunction...

My complaint is more that the output of a single sub-agent is always too much, by its nature it's basically a glorified sub-Claude chat which can get quite long and working with it in the terminal is tedious to simply breaking / unusable.

This is a bug in the sense my terminal becomes unusable with the current setup to interact with sub-agent output. It's a Feature Request in the sense I'm providing some suggestions on how this could be enhanced / fix for better user experience.

Thank you, keep up the good work!

What Should Happen?

I'd like to get granular details of the sub-agents to understand what they're doing, what tools they're using, what files being updated, etc. would like this as a summary of some sort in clean form after each sub-agent's used (not just "27 tools used, this many tokens, took this long of time, ..." - not all that helpful).

Also want to follow along better with what changes are being done as they're being done... Even if it's just a means to "watch tail" the last 20 lines or something (with any tool uses like file changes done in summary means with ability to expand)... Something like that, clean, organized, similar experience as full agent, but baked into the sub-agent in a convenient manner (I often read and follow along with Claude's output, so only the last few thoughts / lines / tool uses are really relevant, if I can do that with the agents, that'd be great as well and solve this strictly for the scenario I'm following along, but doesn't fix it for when I want to refer back to what a sub-agent did...).

Error Messages/Logs

Steps to Reproduce

1. Open Claude Code where sub-agents exist
2. Do a task requiring sub-agents
3. Observe how the sub-agents do not provide much meaningful output while they're working. When "ctrl+r" to try and see more, I'm greeted with either a "ctrl+e" screen (which once I "ctrl+e" same result I mention next happens), or a flickering terminal, blank terminal, etc. oftentimes breaking it so I have to kill the terminal and relaunch Claude Code altogether... -- very disruptive... I'm unable to follow along with sub-agents in a meaningful way or review work it has done in hindsight...

Claude Model

None

Is this a regression?

No, this never worked

Last Working Version

_No response_

Claude Code Version

1.0.83 (Claude Code)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

all terminals this applies to...