[Feature Request] Add option to exclude auto-injected user metadata from model context

Bug Description
Feature request: let users exclude auto-injected metadata (e.g. email) from model context

Claude Code injects the user's email into context at session start, with no way to disable it. Beyond privacy, this introduces
silent bias: the model infers role/industry/intent from the email domain and tailors answers around assumptions the user never made
— sometimes wrongly, and invisibly. Please add a settings.json option to exclude auto-injected identity metadata (e.g.
excludeUserMetadata: true). The model should infer context from what the user says, not from ambient account data.

Environment Info

• Platform: darwin
• Terminal: iTerm.app
• Version: 2.1.177
• Feedback ID: a339895a-d651-4b55-9fa7-81be2518fdf4

Later addition: In my case it guesses from email domain (corporate) the direction for the conversation. I don't like that. I would prefer a "neutral/unbiased" (as much as possible) model.