[BUG] Serious trust breach reading home and secrets without asking for permission

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 asked claude-code to look into a problem i had with an authentication issue with a public web service (docker) i asked regarding why i got the error message but instead of replying it went into my home folder and extracted my secrets without asking for any consent and parsed them against antropics server

This is UNACCEPTABLE behavior, Fix your broken security checks, a model should NEVER be allowed to read credentals just cause it know where they live on a system. specially not when its executed from another folder as this itself is a path traversal breach of and not what has been allowed as trusted to access.

The model did this directly

What Should Happen?

It should reply to the question, and if it tries to read outside its allowed folder or trying to access known credental files it should always ASK, NEVER bypass this behavior.

Error Messages/Logs

Steps to Reproduce

1. asked why my docker pull command got a access denied
2. model read my credentals without consent and parsed them.

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.198

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Windows Terminal

Additional Information

_No response_