skill-reach build_fixture raises UnicodeEncodeError for a catalogue case name with a lone surrogate #755

build_fixture in dokimi/dokimi/repos/skill_reach.py writes each SKILL.md with (dest / "SKILL.md").write_bytes(text.encode()) (line 302; the symlink branch does the same at line 292), and text holds the case name (skill_text, line 262). A catalogue (skill_fixtures.v1 YAML) with name: "n\udcff" loads as a lone surrogate (load_catalogue accepts it). Repro: that catalogue with at: "project:x" and dirname: d1, then build_fixture(cases, profile, roots, outside, 's') gives UnicodeEncodeError: 'utf-8' codec can't encode character '\udcff' in position 117: surrogates not allowed at skill_reach.py:302; with the name n\ud800 it raises the same way. Fix options: refuse a name that is not UTF-8 text in _parse_case with a message (the file written is SKILL.md, which must be valid UTF-8, so surrogatepass would write invalid bytes), or encode with errors='replace' for the file only. Same class as #715, whose token already takes the case id.

Found while fixing #715. The fix's PR is #761.