Copying Agent responses corrupts Unicode chars (…, —, → become mojibake)

[jonaswebdev]

Bug

Copying Agent response text out of Switchboard corrupts certain Unicode characters. Clipboard content shows mojibake instead of the actual glyph.

Observed mapping

Expected	Got (clipboard)
… (ellipsis, U+2026)	‚Ħ
— (em dash, U+2014)	‚Äî
→ (right arrow, U+2192)	‚Üí

Likely cause

Pattern matches double-encoding mojibake: UTF-8 bytes for these chars get misread as Windows-1252/Latin-1 somewhere in the copy pipeline, then re-encoded as UTF-8 on top of that.

Example: … = UTF-8 E2 80 A6. Read as CP1252 → …. Re-encode that string as UTF-8 → ‚Ħ. Same transform reproduces the — and → cases.

Suspect a spot in the copy handler (or a lib it calls) that does a Latin-1/CP1252 decode on already-UTF-8 bytes before writing to clipboard.

Repro

1. Get an Agent response containing …, —, or →.
2. Copy it (button or manual select+copy — note which).
3. Paste into plain-text editor.
4. Chars show as ‚Ħ, —, ‚Üí instead of originals.

Expected

Clipboard content matches on-screen text exactly.

[jonaswebdev]

It was fixed in the 0.0.33/34 releases, thank you