Zed context auto injected to agent prompts

[thezachdrake]

The models are just so convinced we are using vscode and they always try to create things or edit things in .vscode when setting up tooling. It would be amazing if there was some user level and system level information injected automatically into the prompts, kind of like the rules are. For example

Dev Environment:
    - Editor: Zed Nightly 0.203.0
    - Directory: /current/path
    - OS: NixOS
    - etc.
    - etc.
User:
    - Name: Zach
    - etc. 
    - etc.

The more the agent is just automatically aware of the environment the better recommendations and decisions it will make.