my regex line {
    ^^ :my $s; (\w+) <?{ $s = %sigils{$0} }> \h+ $s (\N*) $s $$
}

.say for "star *foo*\nat @bar\nat @baz@" ~~ m:g/<line>/;

That's HOT! :-D

But how would I rewrite my tests, then — which actually want to operate on the captured value?

This seems clunky and naïve: (Not unlike myself... «grin»)

#!/usr/bin/env perl6
use Test;

my %sigils =
    bang   => '!', at      => '@', hash  => '#',
    dollar => '$', percent => '%', caret => '^',
    and    => '&', star    => '*', zero  => '0';

my regex line {
    ^^ :my $s; (\w+) <?{ $s = %sigils{$0} }> \h+ $s (\N*) $s $$
}

is  ('bang !one!' ~~ m/<line>/)<line>[1],     'one', 'is   bang!one';
is  ('zero 0one0' ~~ m/<line>/)<line>[1],     'one', 'is   zero0one';
isnt 'bang @one@' ~~ m/<line>/ && $<line>[1], 'one', 'isnt bang@one';
isnt 'BONK !one!' ~~ m/<line>/ && $<line>[1], 'one', 'isnt BONK!one';
nok  'bang @one@' ~~ m/<line>/, 'mismatch fails';
nok  'BONK !one!' ~~ m/<line>/, 'unknown fails';
[download]

But how would I rewrite my tests, then — which actually want to operate on the captured value?

This is Perl, so there's more than one way to do it, eg

my %sigils =
    bang   => '!', at      => '@', hash  => '#',
    dollar => '$', percent => '%', caret => '^',
    and    => '&', star    => '*', zero  => '0';

my regex line {
    ^^ :my $s; (\w+) <?{ $s = %sigils{$0} }> \h+ $s (\N*) $s $$
    { make ~$0 => ~$1 }
}

sub parse($_) { map *.<line>.made, m:g/<line>/ }

say parse "star *foo*\nat @bar\nat @baz@";
[download]

will generate a sequence of pairs with the key being the name of the sigil and the value the enclosed text.

You can name a capture if you think that's an improvement over numbered captures (see the two lines with comments):

my regex line {
    ^^
    (\w+) :my $s; <?{ $s = %sigils{$0} }>
    \h+
    $s $<foo>=\N* $s # Note the foo bit here ...
    $$
}

is  ('bang !one!' ~~ m/<line>/)<line><foo>, 'one', 'is   bang!one'; # 
+... and here
[download]

Another tweak is to drop one level of hash keys by using a grammar (again, see the two lines with comments):

grammar g {
    regex TOP { # drop 'my'; name 'TOP' to simplify .parse call in `is
+` test
    ^^
    (\w+) :my $s;  <?{ $s = %sigils{$0} }>
    \h+
    $s $<foo>=\N* $s
    $$
    }
}

is  g.parse('bang !one!')<foo>,     'one', 'is   bang!one'; # calling 
+.parse method on grammar defaults to starting with rule (regex) named
+ 'TOP'
[download]

perl6.party (not mine)