Special characters

Everything in a visible term or an entry= level is literal text. You never escape for LaTeX yourself — the extension does it, including for characters that would otherwise break the build or act as index operators. Every printable ASCII character works in both places, including an unbalanced brace. examples/escaping.qmd puts each one in an index entry on its own, and examples/xref-escaping.qmd puts each one in a cross-reference target — a harder place, since a target travels through the index tool’s encapsulation channel, where an unquoted ! is rejected outright and the whole render fails. The test suite compiles both with the same engine your PDF build uses and checks that the index tool accepted every entry. It additionally confirms that the sixteen characters needing special handling — % & # _ { } \ ~ ^ $ @ | ! " < > — actually typeset, in an entry and in a cross-reference alike.

Two characters need care in entry="…", because Quarto’s markdown parser consumes one level of backslash escaping in a quoted attribute before the extension ever sees the value:

To get	Write
a literal \	\\
a literal "	\"
a literal !	!!

Note that \! is not an escape. The parser turns it into a plain !, which the extension then reads as a level separator — so entry="A\!B" produces the entry A with sub-entry B, not A!B. Use !!.

The same two rules apply inside see= and see-also=, and inside sort=.

@, | and " are ordinary literal characters here. Locator styling, which uses some of them in raw makeindex syntax, is not part of this syntax and will arrive later as a separate span attribute.

All of that is about ASCII. A term written in another script needs one more thing from your document — see Terms outside Latin-1.