Your sample code does not address the problem.

The arguments are those passed to the lvalue sub that define what lvalue is returned by the sub, and therefore what gets modified by the assignment.

In order to validate that part of the data that was modified, you need to know those arguments. So, show me how you can avail the FETCH and STORE methods of the tie of those arguments, so that they can achieve the same validation as is shown in the commented out code in the sub modifySubstring()?

#!/usr/bin/perl
use strict;
use warnings;

package UglyTie;
use Data::Dumper;

sub TIESCALAR {
    my $class = shift;
    bless [ @_ ], $class;
}

sub STORE {
    warn Dumper @_;
}

sub FETCH {
    warn Dumper @_;
    1;
}
package SillyClass;

sub new {
    bless \$_[ 1 ], $_[ 0 ];
}

sub modifySubstring: lvalue {
    my( $self, $start, $length ) = @_;
    substr( $$self, $start, $length );

      ## Validate that part of the string that was modified
    # die 'Bad value' 
    #    unless substr( $$self, $start, $length ) =~ m[^[a-z]+$];
}

1;

package main;

my $silly = SillyClass->new( 'a teststring' );
$silly->modifySubstring( 3, 3 ) = 'ABC';
[download]

Same problem here. How can I perform the validation of the values assigned to the slice, as shown commented out, through a tie, or traits, or a type definition?

#!/usr/bin/perl
use strict;
use warnings;

package SillyClass2;

sub new {
    my $class = shift;
    bless \@_, $class;
}

sub modifySubset: lvalue {
    my( $self, $start, $end ) = @_;
    
    @{ $self }[ $start .. $end ]
    
    ## Validate that all values in the slice being assigned are
    ## non-zero, even and uniq

    # my %uniq;
    # die 'Bad value' 
    #    unless $end - $start == grep{ 
    #    $_ && !( $_ & 1 ) && not ++$seen{ $_ } 
    #} @{ $self }[ $start .. $end ];
}

1;

package main;

my $silly = SillyClass2->new( map $_ * 2, 1 .. 50 );
$silly->modifySubset( 15, 35 ) = ( 1 .. 20 ); ## should cause an error
+!
[download]