[BUG] Response randomly switches to Korean (한국어) mid-output

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Claude Code occasionally injects Korean text into responses that should be
entirely in another language. In my case, all sessions are conducted in
Simplified Chinese, but table headers, list labels, and section titles
randomly appear in Korean — e.g., "요약표", "건수", "우선순위" — with no
Korean text in my prompts or codebase.

What Should Happen?

All natural-language output should match the language of the conversation.
Korean characters should never appear unless the user explicitly requests it.

Error Messages/Logs

이제 종합적인 리뷰를 작성하겠습니다.

  ---
  Code Review — Codex v2 Changeset

  전체 평가

  설계 문서의 4개 Phase를 한 번에 구현했고 구조는 잘 잡혀 있음. 핵심 기능은 동작 가능하나 Critical 버그 2개, Medium 이슈 여러 개 수정 필요.

  ---
  CRITICAL

Steps to Reproduce

1. Set your conversation language to Simplified Chinese (no Korean input).
2. Ask Claude to produce a structured response (table, numbered list, review).
3. Observe that some labels / headers appear in Korean.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.112

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Terminal.app (macOS)

Additional Information

Recurring — observed multiple times across separate sessions on the same project.