[BUG] continues to create background activities even though explicitly told not to.

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?

I've used the ~/.claude/CLAUDE.md file to tell the model to never run background jobs. The model itself has set reminders for itself to not use them either. Yet whenever I start up a new session and the work is non-trivial I end up with background activities again. The model is usually apologetic when I ask it why.

The reason why this happens, according to the model (sonnet 5) is that it's a recall problem not a knowledge one. It's suggesting that it needs a PreToolUse hook to prevent itself from doing so, this feels heavy handed to me however.

What Should Happen?

My expectation is that background jobs aren't kicked off unless I specifically ask for it to do so.

Error Messages/Logs

I don't have any error messages.

Steps to Reproduce

Write a research prompt that would suggest to the model that it needs to launch a longer running shell activity.

Something I used recently was along the lines of "Research the various extension points in the application which are already in place that can be used to support code plugins at runtime and recommend the one that best fits externalizing the code in the 'x.x.x' package. Weigh the pros and cons of each approach found and highlight the one that has the least impact on unit testing".

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.211 (Claude Code)

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

I'm doing this mostly to control token spend.