I just visually pair them. The code between the tokens is usually short enough that matching them is not an issue. I remember having some issues matching them on a code that was a little bit more complex but I eventually simplified it so the issue was gone.

This is the most “complex” code I have written with it:

<?xml version='1.0' encoding='utf-8'?>
<feed xmlns='http://www.w3.org/2005/Atom'>
<id>https://adi.tilde.institute/</id>
<title>Adi, Freelance Developer</title>
#!
for f in *
do
test -d "$f" && continue
case $f in
	*.html|*.xml|*.txt|*.css|*.js|*.ico|*.png|*.upp*|feed|urlenc)
        continue
        ;;
esac
stat -f '%Sm' -t '%Y-%m-%dT%H:%M:%SZ' "$f"
done | sort -r | head | {
read -r u
#!
<updated>$u</updated>
#!
}
#!
<author>
<name>Adrian Emil Grigore</name>
<email>adi@tilde.institute</email>
<uri>https://adi.tilde.institute</uri>
</author>
<link href='https://adi.tilde.institute/atom.xml' rel='self' />
#!
for f in *
do
test -d "$f" && continue
case $f in
	*.html|*.xml|*.txt|*.css|*.js|*.ico|*.png|*.upp*|feed|urlenc)
        continue
        ;;
esac
#!
<entry>
	#!
	echo "$f" | ./urlenc | {
	read -r ef
	#!
	<id>https://adi.tilde.institute/$ef.html</id>
	<link href='https://adi.tilde.institute/$ef.html'/>
	#!
	}
	#!
	<title>${f#* }</title>
	#!
	stat -f '%Sm' -t '%Y-%m-%dT%H:%M:%SZ' "$f" | {
		read -r u
	#!
	<updated>$u</updated>
	#!
	}
	#!
	<summary>
	#!
	tail -n +4 "$f" | head -4 | sed 's/^/	/'
	#!
	</summary>
</entry>
#!
done
#!
</feed>