[BUG] Unable to get Claude Code to condense response in plan mode

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?

When I am in plan mode and iterative with Claude Code on a plan, I specifically tell it to condense the response (sometimes as explicit as "reduce response length by 50%, or do NOT respond with the overall plan every time), but it almost always repeats the entire plan and spams the chat, making it extremely unclear how far to scroll back up to read the response. The response is very spam-like and aims to clearly outline what it's doing and more, but encourages users to just ignore the plan as it is extremely difficult to figure out how far up to scroll and where to start reading, and there is a ton of repeated material (e.g. when iterating and asking for small changes to the plan, each time it repeats the ENTIRE plan which may be hundreds of lines long). There isn't a foolproof way or any way to get it to ONLY repeat the new changes. I have tried explicitly telling it to only repeat the new changes and do NOT repeat the entire plan, and adding this to claude.md, but about 90% of the time it still spams the chat.

What Should Happen?

Claude needs to not spam the chat and be concise, and respect user requests when explicitly told not to spam and repeat the entire plan, either thru the chat or in claude.md but it mostly does not listen.

Error Messages/Logs

Steps to Reproduce

Ask claude to do something in plan mode, then ask it for minor adjustments multiple times, 90% of the time it will spam the chat every time.

Claude Model

Not sure / Multiple models

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

latest

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

VS Code integrated terminal

Additional Information

_No response_