Feature request: user-editable global memory for Codex preferences

Feature request

Add a user-editable global memory/preferences surface for Codex that works across repositories and sessions.

Problem

Some recurring workflow preferences are not repo-specific, but they matter every time Codex works with a user. Today those preferences have to be repeated in each thread or written into each repository, which is not practical when working across many repos.

Example: for one organization, stage deployment verification is normally done via the service /version endpoint because Kubernetes access is usually unavailable. That convention applies across many services, but it does not belong in every repository README.

Proposed behavior

Codex should provide a global memory/preferences area that the user can view and edit. Codex could consult it during task execution, similar to ChatGPT memory, but optimized for coding workflows.

Useful properties:

• User-visible and user-editable.
• Applies across repos and future sessions.
• Supports organization/repo/domain scoping where possible.
• Lets users store durable workflow preferences, not secrets.
• Easy to disable, inspect, correct, or delete.

Example memories

• For x services, verify stage deployments via the /version endpoint; Kubernetes access is usually unavailable.
• Prefer feature/<ticket>-... branch names for work repos.
• When asked to combine dependency updates, use a generic commit message and avoid noisy provenance in the commit text.

Why this matters

Repo-local notes do not scale for users who work across many repositories. A global Codex memory would reduce repeated instructions, make agent behavior more consistent, and keep personal workflow preferences out of source-controlled project documentation.