+my $help = <<EOF;
+usage: public-inbox-learn [OPTIONS] [spam|ham|rm] </path/to/RFC2822_message
+
+required action argument:
+
+ spam unindex the message and train as spam
+ rm remove the message without training as spam
+ ham index the message (based on To:/Cc: headers) and train as ham
+
+options:
+
+ --all scan all inboxes on `rm'
+
+See public-inbox-learn(1) man page for full documentation.
+EOF