[BUG] - Endless loop and excessive tool calls by sub agents contributing to massive spike in token consumption (Max plan user)

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?

Claude code is using sub agents without me asking it to and on a very simple repository of around 500 lines of code it is using sub agents to explore the repo and they are caught in recursive loops. They are also using an astounding 106 tool calls on an insanely basic prompt to just explore the 500 line repository. The sub agent never comes back with a report of its findings either. See the screenshot.

What Should Happen?

Sub agents should not be using 106 tool calls to explore a 500 line repository and should not be timing out. This is a simple task.

Error Messages/Logs

Steps to Reproduce

1. Be on max 5x plan, latest Claude code cli version
2. Provide your GitHub repo URL
3. Ask Claude to explore the repo and get an understanding by using sub agents.
4. explore sub agents will use excessive tool calls and not come back with a report to the main agent.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.86

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

VS Code integrated terminal

Additional Information

!Image
!Image