I've got some wonderful new rules that would help classify mail messages! How do I contribute them?

Ideally you will have tested your rules against your own sets of (hand-checked) spam and non-spam messages to be sure they do a good job of identifying messages correctly. This doesn't need to be 100% accurate - but does need to be fairly close to 100%.

Not checking the rules against non-spam is a very bad idea; we've often found that rules that seemed like a great idea, accidentally hit a lot of non-spam in ways we hadn't foreseen.

Next, you should join the spamassassin-users mailing list (if you're not already a member), and post the test rules to that list, and ask others to check against their corpora. List members often will also suggest improvements to the tests.

Most importantly, you MUST create an entry with the suggested rules in the SpamAssassin "bugzilla" system. This will ensure that the developers will actually see the rules (while the developers participate in the list, it's easy for these suggestions to get lost in the noise). You should attach 1 or more sample spam messages (with all headers intact, please!) as attachments, too, so we can see what the rule is intending to catch.

New rules are added to the SVN (development) repository for testing, and if they work well, are likely to be included in future released versions.

You might also post your SpamAssassinRules to http://www.exit0.us .

You can also post your SpamAssassinRules to the SARE forums at http://www.rulesemporium.com/forums/ – SARE will test rules that seem promising, and can often incorporate viable rules into their custom files faster than such rules get added to the distribution set.