Spam Wave Rule Tuning Agent

A new spam wave uses misspelled links, and at the same time the current rules remove a member's genuine post about a sale. This agent reads recent removals, user reports and appeals, then builds candidate rules to catch the new pattern, such as a phrase, a link pattern or a new-account posting rate. It tests each rule against historical posts, both known spam and known good posts, and measures how many spam posts it catches and how many good posts it would wrongly hit. It adjusts the rule and retests until the balance meets the targets. It also checks that a rule does not conflict with existing rules. The moderator approves the rule going live. Edge case: a rule that would catch 99 percent of spam but hit 4 percent of good posts is revised, not approved.