[gal-a]

On Windows, in Python's default mode, running markdown-code-runner on a
markdown file crashes and leaves the file empty.

The file is opened for writing with the locale's encoding, which is cp1252
on most Western Windows setups. Opening it for writing empties it, and then writing the
⚠️ This content is auto-generated marker fails, since cp1252 has no ⚠️.
The output file is the input file by default, so the original is gone.

To check, in Git Bash on Windows with Python 3.9 to 3.14 and without
PYTHONUTF8 set:

printf 'Some text\n\n<!-- CODE:START -->\n<!-- print("hi") -->\n<!-- CODE:END -->\n<!-- OUTPUT:START -->\n<!-- OUTPUT:END -->\n' > demo.md
python -m markdown_code_runner demo.md

Before: UnicodeEncodeError, and demo.md is 0 bytes. After: the output is
written and the file keeps its text.

This passes encoding="utf-8" to the four reads and writes in
markdown_code_runner.py. CI runs on Ubuntu, where UTF-8 is the default,
which is why it did not show there.

No test added: on Ubuntu a test would pass with or without the fix.

Found with qikly, an AI-assisted test tool
I maintain, when the project's own tests failed on Windows. I reproduced it
by hand, and reviewed and tested this change myself.

[greptile-apps]

Confidence Score: 5/5

[Medium impact] This PR appears safe to merge.

Summary

The PR makes all four file reads and writes in markdown_code_runner.py explicitly use UTF-8.

• The runner reads and writes its files as UTF-8.

_{Reviews (1) · Last reviewed commit: "Read and write files as UTF-8" · Reviewed by Greptile}