Projects should remain isolated without shared memories

Environment

• Platform (select one):
• [x] Anthropic API
• [ ] AWS Bedrock
• [ ] Google Vertex AI
• [ ] Other: <!-- specify -->
• Claude CLI version: 1.0.103 (Claude Code)
• Operating System: EndeavourOS-Latest (Arch)
• Terminal: Kitty

Bug Description

When asked to share my information, I definitely selected the second option, "2. Accept Terms - Help improve CLaude: OFF". I selected this option because I don't want to share my clients' data with a persistent database that is out of my control (GDPR).

When I started a new project this morning and ran the traditional, cluade /init, I was shocked to see it referencing another project in a completely different directory. I can't go into details due to privacy reasons, but claude was actually trying to deal with a previous issue in a completely different project. That previous project has a completely different scope from the one that I'm currently in. There are no "similarities" at all.

It doesn't appear this is an issue with history items in the .claude.json file in my user root directory bleeding into other projects. It appears that the "memory" functionality is being saved remotely (I hope I'm wrong).

In any case, Claude was trying to apply fixes to the current project using data from another project, when I I was trying to do was run a /init command. I litterally had to pause Claude code and tell it, "I just want a claude init. I have other work in this project to do. Please just write a basic CLAUDE.md evaluation."

Steps to Reproduce

1. Spend a few hours wtih one project
2. Start a second one and run claude /init

Expected Behavior

I don't expect the data of the separate project to be "Linked" in any sort of fashion. I don't expect the code of my projects or processe to be stored remotely. Every project should be self-contained and Isolated.

Actual Behavior

Multiple projects' data is being shared and transmuted.

Additional Context