[Codex Security] Workspace fails on Japanese Windows when Git commit subject contains non-ASCII characters

What issue are you seeing?

When attempting to open the Codex Security workspace, workspace creation fails while inspecting the target repository’s Git metadata.

On Japanese Windows, opening a Codex Security workspace fails when the repository’s HEAD commit subject contains Japanese characters.

Codex Security runs Git through Python with text=True but without an explicit encoding. Git returns the commit subject as UTF-8, while Python decodes it using the Windows default encoding, CP932.

This first causes a UnicodeDecodeError. Because the subprocess reader then leaves stdout as None, a secondary AttributeError occurs.

Environment:
Windows 11 Pro, Japanese locale
Python preferred encoding: cp932
Codex Version 26.616.81150 • Released 2026/06/24
Codex Security plugin: 0.1.10
Example HEAD subject: docs: おすすめ機能のアルゴリズム仕様書を更新

Full error:

Exception in thread Thread-11 (_readerthread):
Traceback (most recent call last):
  File "C:\Users\<USER>\.cache\codex-runtimes\codex-primary-runtime\dependencies\python\Lib\threading.py", line 1075, in _bootstrap_inner
    self.run()
  File "C:\Users\<USER>\.cache\codex-runtimes\codex-primary-runtime\dependencies\python\Lib\threading.py", line 1012, in run
    self._target(*self._args, **self._kwargs)
  File "C:\Users\<USER>\.cache\codex-runtimes\codex-primary-runtime\dependencies\python\Lib\subprocess.py", line 1599, in _readerthread
    buffer.append(fh.read())
                  ^^^^^^^^^
UnicodeDecodeError: 'cp932' codec can't decode byte 0x83 in position 22: illegal multibyte sequence

Traceback (most recent call last):
  File "C:\Users\<USER>\.codex\plugins\cache\openai-curated-remote\codex-security\0.1.10\scripts\workbench_db.py", line 3643, in <module>
    main()
  File "C:\Users\<USER>\.codex\plugins\cache\openai-curated-remote\codex-security\0.1.10\scripts\workbench_db.py", line 3580, in main
    result = create_workspace(connection, args)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\<USER>\.codex\plugins\cache\openai-curated-remote\codex-security\0.1.10\scripts\workbench_db.py", line 1164, in create_workspace
    inspected = inspect_setup_values(
                ^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\<USER>\.codex\plugins\cache\openai-curated-remote\codex-security\0.1.10\scripts\workbench_db.py", line 873, in inspect_setup_values
    require_scannable_target(target)
  File "C:\Users\<USER>\.codex\plugins\cache\openai-curated-remote\codex-security\0.1.10\scripts\workbench_db.py", line 935, in require_scannable_target
    metadata = git_target_metadata(target)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\<USER>\.codex\plugins\cache\openai-curated-remote\codex-security\0.1.10\scripts\workbench_target.py", line 441, in git_target_metadata
    "commitSubject": git_output(target, "show", "-s", "--format=%s", "HEAD"),
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\<USER>\.codex\plugins\cache\openai-curated-remote\codex-security\0.1.10\scripts\workbench_target.py", line 27, in git_output
    output = completed.stdout.strip()
             ^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'strip'

What steps can reproduce the bug?

1. Use Windows with a Japanese locale where Python’s preferred encoding is CP932.
2. Create a Git repository.
3. Create a commit whose subject contains Japanese characters.
4. Run a Codex Security scan and attempt to open its workspace.
5. Workspace creation fails while reading the HEAD commit subject.

Example PowerShell reproduction:

git init codex-security-encoding-repro
Set-Location codex-security-encoding-repro

Set-Content README.md "# Test"
git add README.md
git commit -m "docs: 日本語のコミットメッセージ"

python -c "import locale; print(locale.getpreferredencoding(False))"
# cp932

python -c "import subprocess; print(subprocess.run(['git', 'show', '-s', '--format=%s', 'HEAD'], capture_output=True, text=True).stdout)"

The final command raises:
UnicodeDecodeError: 'cp932' codec can't decode byte ...: illegal multibyte sequence

What is the expected behavior?

Codex Security should successfully open and scan repositories whose Git metadata contains non-ASCII characters, regardless of the Windows system locale.

Additional information

This appears related to #14311, which reports the same implicit text=True encoding problem in a different Codex Python component.