ASCII prompt guidance should allow correct non-English orthography

Summary

Codex prompt guidance currently tells the agent to default to ASCII and only introduce non-ASCII or Unicode characters when there is a clear justification and the file already uses them. That is too restrictive for ordinary language-writing tasks where Unicode is required for correct orthography.

For example, when a user asks Codex to write text in Spanish, accents and characters such as ñ are not optional styling; they are part of correct spelling. The current wording can lead the agent to incorrectly avoid those characters unless the target file already contains Unicode.

Proposed change

Clarify that correct orthography in a language that requires non-ASCII characters is a valid justification for using Unicode, even if the target file does not already contain Unicode.

Suggested wording:

Default to ASCII when editing or creating files. Only introduce non-ASCII or other Unicode characters when there is a clear justification, such as correct orthography in a language that requires them, or when the file already uses them.

Why this matters

The ASCII default is useful for source files and simple code edits, but it should not override correct natural-language output. This comes up often for non-English text, including Spanish, where removing accents changes spelling and quality.

Possible fix

I prepared a branch with the prompt-only change here:

• Branch: jamarju/codex@codex/allow-unicode-orthography
• Commit: 11875849bd Clarify Unicode prompt guidance
• Compare: https://github.com/openai/codex/compare/main...jamarju:codex/allow-unicode-orthography?expand=1

That branch updates the checked-in Codex prompt sources and embedded model instruction data consistently.

Validation on the branch

• jq empty codex-rs/models-manager/models.json
• Verified the old restrictive sentence no longer appears in the touched prompt sources.