Codex.app leaves behind 1.2 GB code_sign_clone directories on every launch

Resolved 💬 1 comment Opened Jun 16, 2026 by polhenarejos Closed Jun 16, 2026

What version of the Codex App are you using (From “About Codex” dialog)?

Version 26.609.71450 • Released Jun 16, 2026

What subscription do you have?

Pro

What platform is your computer?

macOS Sequoia

What issue are you seeing?

no-active-thread-019ed03e-2d38-7aa0-a177-6b2a13900b09

Environment

  • macOS 15.7.2 (Apple Silicon)
  • Codex for macOS version 149.0.7827.54

Issue

Each time I launch Codex.app, a new directory is created under:

/private/var/folders/.../X/com.openai.codex.code_sign_clone/

The directory contains a full copy of the application bundle:

Codex.app.bundle

The size of each clone is approximately 1.2 GB, which matches the size of the installed application:

du -sh /Applications/Codex.app
1.2G /Applications/Codex.app

The problem is that these directories are not removed when Codex exits. Every launch creates a new clone with a different name, for example:

code_sign_clone.5IQBPt
code_sign_clone.FJSOFT
code_sign_clone.6NGelH

Each of them occupies about 1.2 GB.

Evidence

Example contents:

/private/var/folders/.../com.openai.codex.code_sign_clone/code_sign_clone.XXXXXX/Codex.app.bundle

Disk usage:

1.2G code_sign_clone.5IQBPt
1.2G code_sign_clone.FJSOFT
1.2G code_sign_clone.6NGelH

The clone size exactly matches the installed application size.

What steps can reproduce the bug?

ls -lah /private/var/folders/.../X/com.openai.codex.code_sign_clone/
close the app
open the app
/private/var/folders/.../X/com.openai.codex.code_sign_clone/

and it continues indefinitely.

What is the expected behavior?

Expected behavior

Temporary code-signing clones should be automatically removed when no longer needed, especially after the application exits.

Additional information

Actual behavior

The clones remain on disk indefinitely and accumulate across launches, consuming large amounts of storage (over 13 GB in my case).

View original on GitHub ↗

This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗