3.2.0 — Server Shield V2 scanning, undo controls, spam fingerprinting

Added

• Server Shield now scans images that arrive via link-preview embeds, closing the "post a bare link" evasion where the scam image was never attached to the message. (#152)
• Server Shield now scans Discord CDN image links pasted as message text, closing the "paste the link instead of attaching" evasion. (#153)
• Server Shield now scans the first frame of GIF images, covering both uploaded .gif files and .gif CDN links. (#154)
• The moderation "new account" age cutoff is now a per-server setting on the Server Shield page (1 to 365 days), and its default is raised from 7 -> 30 days to better catch throwaway scam accounts. (#150)
• Moderation flag verdicts are now reversible: after confirming or dismissing a flagged image, an "Undo verdict" button lets staff re-decide. (#157)
• Removal mod-logs now carry an "Undo punishment" button that reverses a reversible action (clears a timeout or lifts a ban) and sends the affected member a best-effort apology DM. Kicks and deleted messages can't be undone. (#158)

Changed

• Landing page refresh: a shared footer with a provenance line, updated hero copy, and a version badge in the header linking to the release notes. (#149)
• Spam attachments are now fingerprinted by file content, so re-posting an identical file under different names across channels is caught instead of slipping past the per-channel threshold. (#155)

Fixed

• The bot no longer responds twice to commands and events while a deploy is in flight. Only one bot process is active at a time; expect a few seconds of unresponsiveness at deploy cutover instead. (#148)
• The Cross-Channel Spam Guard now shows why a save failed instead of silently rejecting an out-of-range value (for example a window over 60 seconds). (#151)