Flagged Content Review Queue Agent

Moderators face a queue of reports where a death threat can sit behind fifty spam flags. This agent reads every new report and the reported post with its context: the thread, the poster's history and the reporter's history. It sorts reports by severity, with threats, self-harm and doxxing first, then harassment, then spam and off-topic. For each item it suggests an action and quotes the exact guideline involved. It then checks the suggestion against similar past decisions. If it differs from precedent, it flags the item for a moderator rather than suggesting. It also checks whether a burst of reports looks coordinated. Moderators decide every action. Edge case: mass reports from accounts that joined at the same time are flagged as possible brigading instead of counted as real reports.