Version of the library

0.64.2 (also master @ f13a96b)

Describe the bug

If a comment sits after a key's colon and before its value, the comment is parsed as the value. The real value is then lost, or it spills into a bogus key and flattens a nested object. Comments elsewhere (before a key, between array items, after a value) are already skipped correctly. LLMs often annotate JSON this way, for example "age": // estimated\n 42.

How to reproduce

from json_repair import repair_json
for s in ['{"a": /* comment */ 1}', '{"a": // note\n "x", "b": 2}', '{"a": /* c */ {"b": 1}}', '{"a": # c\n true}']:
    print(repr(s), "->", repair_json(s))

Output:

'{"a": /* comment */ 1}' -> {"a": ""}
'{"a": // note\n "x", "b": 2}' -> {"a": "", "x\", \"b": 2}
'{"a": /* c */ {"b": 1}}' -> {"a": "", "b": 1}
'{"a": # c\n true}' -> {"a": ""}

Expected behavior

The comment is skipped and the value is kept: {"a": 1}, {"a": "x", "b": 2}, {"a": {"b": 1}} and {"a": true}.

I have a small fix with tests ready (it skips comments in _parse_object_value before dispatching the value) and can open a PR if that works for you.