How do I instruct the codex agent to edit files using CP850 encoding? Every time I request a task, it edits the files using UTF8, and it bugs out several characters in my source code.
The best way would be to dig into the agent code, and use a robust code library to do this conversion.
Then deal with what the AI can still write that is unsupported in under 230 characters of ANSI code pages.
Asking nicely isn’t going to make the language AI produce upper-bytes of ASCII reliably for the dozen different possibilities for 20-year-old international files. It is essentially a UTF-8 model.