You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Added
Owner-curated global scam-image blocklist. A right-click "Toggle global scam block" command lets the bot owner mark an image's fingerprint as a known scam across every server, without shipping a build. Matches are treated like a normally-detected scam: each server's own dismissal still wins, and the owner's list never applies the harsher confirmed-punishment escalation reserved for a server's own confirmations. (#160)
Welcome messages gain a per-server "Ping on join" toggle, so the member mention can still render as a clickable pill without firing a notification. On by default, so existing servers keep pinging. (#166)
Open Graph and Twitter card meta tags, so a shared shrkbot link shows a proper title, description, and preview image. (#164)
Changed
Single-selection role menus can now toggle a role off: clicking the role you already hold removes it without replacement, instead of doing nothing, so you can return to holding none of the set. (#167)
The manual "Report as scam" and global-scam commands now inspect link-preview embeds and pasted CDN image links, not just attachments, matching the automatic scan. The per-source image cap is also raised from 3 -> 4. (#163)
Confirmed scam-image fingerprints are now retained for 180 days instead of 30, so a moderator's confirmation survives a scam campaign going quiet for a month or more. (#171)
The "Report as scam" action is now surfaced as an info callout under the consent warning on the Image Scanning page, instead of a small footer note. (#170)
Scam-text keyword weights retuned for the current scam wave: added crypto-casino and bonus-offer terms, and lowered "withdraw" so it no longer flags on its own in legitimate messages. (#171)
Fixed
shrkbot no longer double-posts its own kicks, bans, and timeouts to the moderation log. (#169)
Toggling image scanning from the Server Shield overview page no longer wipes the confirmed-punishment settings on a config whose image-scanning options were never edited. (#165)
The public site header logo now links to the home page. (#168)