ВЕРСИИ (82)

## ScamScreener 2.6.1 This release brings back Minecraft `26.1.x` support next to `26.2.x` and trims a few chat false positives. - added Minecraft `26.1.x` builds again through a multi-version build setup - lowered the default too-good-to-be-true score so a lone `guaranteed` or `dupe` mention no longer opens a review alert by itself - stopped treating casual standalone `vc` mentions as external-platform pushes (existing rules files are migrated automatically) - removed the unused legacy trend-escalation mapping and settings - reposted Skyblocker helper lines (score, crypt, and livid callouts) stay out of the trend wave; the recent reports came from clients older than 2.5, so please update - updated build dependencies (Fabric API, Mod Menu, Gradle, Stonecutter)

## ScamScreener 2.6.1 This release brings back Minecraft `26.1.x` support next to `26.2.x` and trims a few chat false positives. - added Minecraft `26.1.x` builds again through a multi-version build setup - lowered the default too-good-to-be-true score so a lone `guaranteed` or `dupe` mention no longer opens a review alert by itself - stopped treating casual standalone `vc` mentions as external-platform pushes (existing rules files are migrated automatically) - removed the unused legacy trend-escalation mapping and settings - reposted Skyblocker helper lines (score, crypt, and livid callouts) stay out of the trend wave; the recent reports came from clients older than 2.5, so please update - updated build dependencies (Fabric API, Mod Menu, Gradle, Stonecutter)

## ScamScreener 2.6.1 This maintenance release updates the build and mod dependencies for both supported Minecraft versions and reduces a chat false positive. - removed the unused legacy trend-escalation mapping and settings - stopped treating casual standalone `vc` mentions as external-platform pushes

## ScamScreener 2.6.1 This maintenance release updates the build and mod dependencies for both supported Minecraft versions and reduces a chat false positive. - removed the unused legacy trend-escalation mapping and settings - stopped treating casual standalone `vc` mentions as external-platform pushes

## ScamScreener 2.6.0 This release adds Minecraft `26.2.x` support and updates the in-game GUI wiring to match the new client API. - added support for Minecraft `26.2.x` - updated ScamScreener screen opening and profiler HUD handling for the `26.2` client API - refined account-data detection so harmless mentions score low while direct credential requests remain high-risk - adjusted cross-sender trend scoring to add one point per other sender, with a fixed `+6` wave score from six participating players - ignored known Discord security-warning messages for link and external-platform scoring while keeping normal Discord invitations active

## ScamScreener 2.6.0 This release adds Minecraft `26.2.x` support and updates the in-game GUI wiring to match the new client API. - added support for Minecraft `26.2.x` - updated ScamScreener screen opening and profiler HUD handling for the `26.2` client API - refined account-data detection so harmless mentions score low while direct credential requests remain high-risk - adjusted cross-sender trend scoring to add one point per other sender, with a fixed `+6` wave score from six participating players - ignored known Discord security-warning messages for link and external-platform scoring while keeping normal Discord invitations active

## ScamScreener 2.5.2 This patch fixes false scam-review cases from reposted mod/helper chat lines by classifying them as `SYSTEM_PLAYER` instead of normal player chat. - reclassified reposted helper callouts from visible player lines as `SYSTEM_PLAYER`, so they keep separate source semantics without entering the normal player-chat scam pipeline - fixed false case creation for wrapped lines such as `[Skyblocker] The livid color is LIME`, `[Skyblocker] 300 Score Reached!`, and `[Skyblocker] We only have 4 crypts out of 5, we need more!`

## ScamScreener 2.5.1 This patch tightens repeat-wave handling and fixes a config regression around the local Training Hub client ID. - tightened `TrendStage` so only repeated scam-like pitches score, instead of generic repeated public chat - moved reposted dungeon/mod helper callouts into one central pre-pipeline player-message filter and added more known lines - preserved the installation-local `trainingClientId` when old or unversioned `runtime.json` files are recreated during schema updates

## ScamScreener 2.5.0 This release retunes the default v2 pipeline to reduce noisy review alerts from common SkyBlock chat traffic. - raised the default review threshold from `1` to `15` - disabled `TrendStage` by default and reduced its repeat-wave fallback tuning - tightened `BehaviorStage` burst defaults - reduced default scores for generic Discord/trust hits and the `external after contact` funnel step - tightened the fuzzy `join my discord` match and reduced the fuzzy `trust me` score - bumped the `RULES` schema version, so existing `rules.json` files are reset to the new defaults on load

## ScamScreener 2.4.3 This patch adds an early pipeline bypass for specific dungeon status messages that some mods repost through player chat, such as `[Skyblocker] 300 Score Reached!`, `Prince dead!`, `Mimic dead!`, and the `[Skyblocker] We only have n crypts out of 5, we need more!` crypt counter lines.

## ScamScreener 2.4.3 This patch adds an early pipeline bypass for specific dungeon status messages that some mods repost through player chat, such as `[Skyblocker] 300 Score Reached!`, `Prince dead!`, `Mimic dead!`, and the `[Skyblocker] We only have n crypts out of 5, we need more!` crypt counter lines.

## ScamScreener 2.4.3 This patch adds an early pipeline bypass for specific dungeon status messages that some mods repost through player chat, such as `[Skyblocker] 300 Score Reached!`, `Prince dead!`, `Mimic dead!`, and the `[Skyblocker] We only have n crypts out of 5, we need more!` crypt counter lines.

## ScamScreener 2.4.2 This patch fixes a chat-classification edge case where system-style mod messages could still be treated as player chat when Fabric exposed sender metadata alongside the visible line.

## ScamScreener 2.4.2 This patch fixes a chat-classification edge case where system-style mod messages could still be treated as player chat when Fabric exposed sender metadata alongside the visible line.

## ScamScreener 2.4.2 This patch fixes a chat-classification edge case where system-style mod messages could still be treated as player chat when Fabric exposed sender metadata alongside the visible line.

## ScamScreener 2.4.1 This update improves the anonymous Training Hub flow inside Minecraft and adds a reminder when you have enough reviewed cases ready to upload. ### Added - Automatic upload reminder once you have more than 5 saved `SAFE` or `RISK` cases. - Clickable `[upload]` and `[don't show again]` actions directly in the reminder chat message. - Review setting toggle to disable the Training Hub upload reminder completely. ### Changed - The Training Hub screen now shows the full local client ID instead of a shortened preview. - Clicking the client ID in the Training Hub now copies it to the clipboard. - Manual uploads started from the reminder or Training Hub now postpone the next reminder for 30 minutes. ### Notes - Upload reminders only appear locally on your client. - The reminder uses the same reviewed `SAFE` and `RISK` case count that the anonymous Training Hub upload exports. - No Microsoft Account Needed. STAY SAFE!

## ScamScreener 2.4.1 This update improves the anonymous Training Hub flow inside Minecraft and adds a reminder when you have enough reviewed cases ready to upload. ### Added - Automatic upload reminder once you have more than 5 saved `SAFE` or `RISK` cases. - Clickable `[upload]` and `[don't show again]` actions directly in the reminder chat message. - Review setting toggle to disable the Training Hub upload reminder completely. ### Changed - The Training Hub screen now shows the full local client ID instead of a shortened preview. - Clicking the client ID in the Training Hub now copies it to the clipboard. - Manual uploads started from the reminder or Training Hub now postpone the next reminder for 30 minutes. ### Notes - Upload reminders only appear locally on your client. - The reminder uses the same reviewed `SAFE` and `RISK` case count that the anonymous Training Hub upload exports. - No Microsoft Account Needed. STAY SAFE!

## ScamScreener 2.4.1 This update improves the anonymous Training Hub flow inside Minecraft and adds a reminder when you have enough reviewed cases ready to upload. ### Added - Automatic upload reminder once you have more than 5 saved `SAFE` or `RISK` cases. - Clickable `[upload]` and `[don't show again]` actions directly in the reminder chat message. - Review setting toggle to disable the Training Hub upload reminder completely. ### Changed - The Training Hub screen now shows the full local client ID instead of a shortened preview. - Clicking the client ID in the Training Hub now copies it to the clipboard. - Manual uploads started from the reminder or Training Hub now postpone the next reminder for 30 minutes. ### Notes - Upload reminders only appear locally on your client. - The reminder uses the same reviewed `SAFE` and `RISK` case count that the anonymous Training Hub upload exports. - No Microsoft Account Needed. STAY SAFE!

## ScamScreener 2.4.0 This release replaces the old Training Hub account login inside the mod with a direct anonymous upload flow based on the local `trainingClientId`. ### Changed - Removed the old in-mod Training Hub username/password login flow. - Training uploads now go directly through the installation-local `trainingClientId`. - The Training Hub screen inside Minecraft is now a simpler upload-only UI. ### Added - SHA-256 upload handshake headers so the Training Hub can verify payload integrity together with the submitted client ID. ### Notes - No ScamScreener web account is needed inside the mod anymore. - Client/account linking is now intended to happen later inside the Training Hub website. - No Microsoft Account Needed. STAY SAFE!

## ScamScreener 2.4.0 This release replaces the old Training Hub account login inside the mod with a direct anonymous upload flow based on the local `trainingClientId`. ### Changed - Removed the old in-mod Training Hub username/password login flow. - Training uploads now go directly through the installation-local `trainingClientId`. - The Training Hub screen inside Minecraft is now a simpler upload-only UI. ### Added - SHA-256 upload handshake headers so the Training Hub can verify payload integrity together with the submitted client ID. ### Notes - No ScamScreener web account is needed inside the mod anymore. - Client/account linking is now intended to happen later inside the Training Hub website. - No Microsoft Account Needed. STAY SAFE!

## ScamScreener 2.4.0 This release replaces the old Training Hub account login inside the mod with a direct anonymous upload flow based on the local `trainingClientId`. ### Changed - Removed the old in-mod Training Hub username/password login flow. - Training uploads now go directly through the installation-local `trainingClientId`. - The Training Hub screen inside Minecraft is now a simpler upload-only UI. ### Added - SHA-256 upload handshake headers so the Training Hub can verify payload integrity together with the submitted client ID. ### Notes - No ScamScreener web account is needed inside the mod anymore. - Client/account linking is now intended to happen later inside the Training Hub website. - No Microsoft Account Needed. STAY SAFE!

# ScamScreener 2.3.4 This patch fixes sender-backed chat classification and avoids repeating the same regex work across rule checks. - Chat callbacks that only expose the sender through Fabric message params now still enter the pipeline as player chat. - Rule and funnel regex checks now reuse cached per-message matches instead of running the same pattern repeatedly.

# ScamScreener 2.3.4 This patch fixes sender-backed chat classification and avoids repeating the same regex work across rule checks. - Chat callbacks that only expose the sender through Fabric message params now still enter the pipeline as player chat. - Rule and funnel regex checks now reuse cached per-message matches instead of running the same pattern repeatedly.

# ScamScreener 2.3.4 This patch fixes sender-backed chat classification and avoids repeating the same regex work across rule checks. - Chat callbacks that only expose the sender through Fabric message params now still enter the pipeline as player chat. - Rule and funnel regex checks now reuse cached per-message matches instead of running the same pattern repeatedly.

# ScamScreener 2.3.3 This patch updates the education help link after the scam reference page moved away from Fandom. - Education follow-up messages now open the new Community Wiki page at `https://hypixelskyblock.minecraft.wiki/w/Scams`.

# ScamScreener 2.3.3 This patch updates the education help link after the scam reference page moved away from Fandom. - Education follow-up messages now open the new Community Wiki page at `https://hypixelskyblock.minecraft.wiki/w/Scams`.

# ScamScreener 2.3.3 This patch updates the education help link after the scam reference page moved away from Fandom. - Education follow-up messages now open the new Community Wiki page at `https://hypixelskyblock.minecraft.wiki/w/Scams`. - Minecraft `26.1`, `26.1.1`, and `26.1.2` now ship as one shared `26.1.x` build instead of separate per-patch artifacts.

# ScamScreener - Blacklist entries added by player name now resolve and persist the player's UUID automatically in the background. - Companion mods querying ScamScreener's runtime-backed API data now receive the current state when calling the API again instead of only the initialization state. - Returned API lists and entry objects are still snapshots, so later changes require a fresh API query.

# ScamScreener - Blacklist entries added by player name now resolve and persist the player's UUID automatically in the background. - Companion mods querying ScamScreener's runtime-backed API data now receive the current state when calling the API again instead of only the initialization state. - Returned API lists and entry objects are still snapshots, so later changes require a fresh API query.

# ScamScreener - Blacklist entries added by player name now resolve and persist the player's UUID automatically in the background. - Companion mods querying ScamScreener's runtime-backed API data now receive the current state when calling the API again instead of only the initialization state. - Returned API lists and entry objects are still snapshots, so later changes require a fresh API query.

# ScamScreener - Blacklist entries added by player name now resolve and persist the player's UUID automatically in the background. - Companion mods querying ScamScreener's runtime-backed API data now receive the current state when calling the API again instead of only the initialization state. - Returned API lists and entry objects are still snapshots, so later changes require a fresh API query.

# ScamScreener - Blacklist entries added by player name now resolve and persist the player's UUID automatically in the background. - Companion mods querying ScamScreener's runtime-backed API data now receive the current state when calling the API again instead of only the initialization state. - Returned API lists and entry objects are still snapshots, so later changes require a fresh API query.

# ScamScreener - Removed the `MarketGuard` hard dependency introduced in `2.3.0`. - `MarketGuard` can now depend on `ScamScreener` again without creating a circular Fabric dependency.

# ScamScreener - Removed the `MarketGuard` hard dependency introduced in `2.3.0`. - `MarketGuard` can now depend on `ScamScreener` again without creating a circular Fabric dependency.

# ScamScreener 2.3.0 - `MarketGuard` is now a required dependency for ScamScreener.

# ScamScreener 2.3.0 - `MarketGuard` is now a required dependency for ScamScreener.

# ScamScreener 2.2.1 This update brings the Training Hub into the mod and cleans up the full contribution flow inside Minecraft. - Integrated the ScamScreener Training Hub directly into the main UI and review flow. - Added in-game ScamScreener account login and upload support for reviewed `SAFE` and `RISK` cases. - Exported training cases now use stable per-installation IDs so repeated exports stay deterministic without tying uploads to the Minecraft account. - Restored the legacy education follow-up messages, including `/scamscreener edu disable <messageId>`. Important: - `1.21.10` is EOS. Please update to `1.21.11` or later.

# ScamScreener 2.2.1 This update brings the Training Hub into the mod and cleans up the full contribution flow inside Minecraft. - Integrated the ScamScreener Training Hub directly into the main UI and review flow. - Added in-game ScamScreener account login and upload support for reviewed `SAFE` and `RISK` cases. - Exported training cases now use stable per-installation IDs so repeated exports stay deterministic without tying uploads to the Minecraft account. - Restored the legacy education follow-up messages, including `/scamscreener edu disable <messageId>`. Important: - `1.21.10` is EOS. Please update to `1.21.11` or later.

# ScamScreener 2.2.1 This update brings the Training Hub directly into the 26.1 mod and cleans up the full contribution flow inside Minecraft. - Integrated the ScamScreener Training Hub directly into the main UI and review flow. - Added in-game ScamScreener account login and upload support for reviewed `SAFE` and `RISK` cases. - Exported training cases now use stable per-installation IDs so repeated exports stay deterministic without tying uploads to the Minecraft account. - Restored the legacy education follow-up messages, including `/scamscreener edu disable <messageId>`.

# ScamScreener 2.2.0 - Minecraft 26.1 with Official Mappings. - Legacy Minecraft Version Support are EOL as of NOW!

# ScamScreener 2.1.3 - 1.21.10 and 1.21.11 EOL Update! - Added a direct `Review Case` button to the `Alert Rule Details` screen, above `Close`. - Warning messages now use `[review]` instead of `[manage]` for the case-review action. - The public `scamscreener-api` pipeline hooks are now live, so external `PipelineContributor` stages are loaded into the runtime pipeline. - Simplified internal config and review plumbing with cleanup of unused wrapper layers.

# ScamScreener 2.1.3 - 1.21.10 EOL Update! - Added a direct `Review Case` button to the `Alert Rule Details` screen, above `Close`. - Warning messages now use `[review]` instead of `[manage]` for the case-review action. - The public `scamscreener-api` pipeline hooks are now live, so external `PipelineContributor` stages are loaded into the runtime pipeline. - Simplified internal config and review plumbing with cleanup of unused wrapper layers.

# ScamScreener 2.1.1 - minor bug fixes

# ScamScreener 2.1.1 - minor bug fixes

# ScamScreener 2.1.1 - added runtime config access for Kd_Gaming1's PackCore Wizard for SkyBlock Enhanced Modpack

# ScamScreener 2.1.1 - added runtime config access for Kd_Gaming1's PackCore Wizard for SkyBlock Enhanced Modpack

# ScamScreener 2.1.0 - added a live in-game profiler HUD with `/ss profiler on` and `/ss profiler off` - added an optional browser profiler through Tango Web API with `/ss profiler open` - web profiler now shows live MSPT/TPS, lifetime averages, phase breakdown, recent event log, reset, and `.sspp` export - profiler can be shared with developers via exported profile files - moved config writes and training export work onto a background file worker to reduce client-thread stalls - reduced message-related micro-lag with a lighter chat ingress path and cached normalization/fingerprint work - senderless mod chat is classified before pipeline entry, and a rare local-echo reentry crash in warning handling is fixed

# ScamScreener 2.1.0 - added a live in-game profiler HUD with `/ss profiler on` and `/ss profiler off` - added an optional browser profiler through Tango Web API with `/ss profiler open` - web profiler now shows live MSPT/TPS, lifetime averages, phase breakdown, recent event log, reset, and `.sspp` export - profiler can be shared with developers via exported profile files - moved config writes and training export work onto a background file worker to reduce client-thread stalls - reduced message-related micro-lag with a lighter chat ingress path and cached normalization/fingerprint work - senderless mod chat is classified before pipeline entry, and a local-echo reentry crash in warning handling is fixed

# ScamScreener 2.0.2 - only validated player messages reach the detection pipeline - system and invalid pseudo-player messages are skipped instead of producing false alerts - added `/ss enable` and `/ss disable` with a clickable re-enable notice on join - added a join-time Modrinth update notification with changelog hover preview - default alert threshold is now `MEDIUM` - config schema versioning is centralized and old or unversioned v2 configs are replaced with current defaults

# ScamScreener 2.0.2 - only validated player messages reach the detection pipeline - system and invalid pseudo-player messages are skipped instead of producing false alerts - added `/ss enable` and `/ss disable` with a clickable re-enable notice on join - added a join-time Modrinth update notification with changelog hover preview - default alert threshold is now `MEDIUM` - config schema versioning is centralized and old or unversioned v2 configs are replaced with current defaults

# ScamScreener 2.0.1 - hardened detection pipeline to ensure system messages are ignored properly

# ScamScreener 2.0.1 - hardened detection pipeline to ensure system messages are ignored properly

# ScamScreener 2.0.0 ScamScreener v2 is a full rewrite of the old v1 line. The release focuses on better scam-signal detection, stronger review workflows, and cleaner training exports. ## Highlights - New modular detection pipeline (rule, similarity, behavior, trend, funnel, context). - New review queue flow with better case handling in GUI. - Canonical export format for training: `training-cases-v2.jsonl`. - Improved command and settings structure under one unified `/scamscreener` command tree. - Broader automated test coverage and release hardening. ## Important For Players Updating From v1 - Legacy v1 whitelist/blacklist is migrated automatically on first v2 startup. - Old v1 config artifacts are cleaned up after migration. - Training Hub is not live yet; `Contribute Training Data` buttons are intentionally disabled. - Case export still works via command: `/scamscreener review export`. ## Command Surface (v2) - `/scamscreener settings` - `/scamscreener review` - `/scamscreener review export` - `/scamscreener whitelist ...` - `/scamscreener blacklist ...` - `/scamscreener rules` - `/scamscreener runtime` - `/scamscreener metrics` - `/scamscreener messages` - `/scamscreener debug ...` ## Notes - This release replaces major parts of v1 internals. - If you run custom tooling around old v1 scripts/paths, update those integrations to the v2 layout.

# ScamScreener 2.0.0 ScamScreener v2 is a full rewrite of the old v1 line. The release focuses on better scam-signal detection, stronger review workflows, and cleaner training exports. ## Highlights - New modular detection pipeline (rule, similarity, behavior, trend, funnel, context). - New review queue flow with better case handling in GUI. - Canonical export format for training: `training-cases-v2.jsonl`. - Improved command and settings structure under one unified `/scamscreener` command tree. - Broader automated test coverage and release hardening. ## Important For Players Updating From v1 - Legacy v1 whitelist/blacklist is migrated automatically on first v2 startup. - Old v1 config artifacts are cleaned up after migration. - Training Hub is not live yet; `Contribute Training Data` buttons are intentionally disabled. - Case export still works via command: `/scamscreener review export`. ## Command Surface (v2) - `/scamscreener settings` - `/scamscreener review` - `/scamscreener review export` - `/scamscreener whitelist ...` - `/scamscreener blacklist ...` - `/scamscreener rules` - `/scamscreener runtime` - `/scamscreener metrics` - `/scamscreener messages` - `/scamscreener debug ...` ## Notes - This release replaces major parts of v1 internals. - If you run custom tooling around old v1 scripts/paths, update those integrations to the v2 layout.

# ScamScreener 1.3.3 Small update in RuleBehavior, lowering false-positive flags.

# ScamScreener 1.3.3 Small update in RuleBehavior, lowering false-positive flags.

# ScamScreener 1.3.3 Small update in RuleBehavior, lowering false-positive flags.

# ScamScreener 1.3.3 Performance and stability update focused on smoother gameplay, lower memory use, and more reliable long-session behavior. ## Fixed - Fixed rare freeze/crash cases triggered by extreme or malformed chat patterns, especially when tying `/guild` Commands - Fixed long-session slowdowns caused by chat/player history growing too much. - Improved overall memory behavior during long play sessions. - Smoothed out internal chat tracking so alerts and review context stay more consistent over time.

# ScamScreener 1.3.3 Performance and stability update focused on smoother gameplay, lower memory use, and more reliable long-session behavior. ## Fixed - Fixed rare freeze/crash cases triggered by extreme or malformed chat patterns, especially when tying `/guild` Commands - Fixed long-session slowdowns caused by chat/player history growing too much. - Improved overall memory behavior during long play sessions. - Smoothed out internal chat tracking so alerts and review context stay more consistent over time.

# ScamScreener 1.3.3 Performance and stability update focused on smoother gameplay, lower memory use, and more reliable long-session behavior. ## Fixed - Fixed rare freeze/crash cases triggered by extreme or malformed chat patterns, especially when tying `/guild` Commands - Fixed long-session slowdowns caused by chat/player history growing too much. - Improved overall memory behavior during long play sessions. - Smoothed out internal chat tracking so alerts and review context stay more consistent over time.

Small quality patch focused on cleaner chat filtering and fewer false captures from non-player system lines.

# ScamScreener 1.3.2 Small quality patch focused on cleaner chat filtering and fewer false captures from non-player system lines.

# ScamScreener 1.3.1 This version keeps the whitelist/review improvements and switches training-data upload to a manual Discord handoff flow. The whitelisted users are ignored by ScamScreener. ## Added - Added whitelist management in-game with both commands and GUI (`/scamscreener whitelist`, `add`, `remove`, `list`, plus `/ss` aliases). - Added a dedicated whitelist config file: `config/scamscreener/scam-screener-whitelist.json`. - Added direct review shortcut commands: `/scamscreener review <playerName>` and `/ss review <playerName>`. - Added a manual upload handoff message that asks to join Discord and includes clickable invite/folder links. ## Changed - Detection now skips whitelist players early in the pipeline (right after mute filtering), before scoring and warnings. - Review auto-marking now follows `AI Auto-Capture` level (`OFF` disables automatic `[S]`; `LOW|MEDIUM|HIGH|CRITICAL` auto-mark only matching-or-higher risk rows). - Whitelist display names are now refreshed toward canonical Mojang names when lookup succeeds. - `/scamscreener upload` now archives training data and prompts manual Discord upload instead of webhook upload. - Upload ToS and README upload/network documentation now reflect player-controlled manual upload behavior. ## Fixed - Fixed review row merge behavior so scored flagged/context rows keep their score signal reliably during review assembly. - Fixed upload status messaging so it no longer implies automatic webhook submission. ## Removed - Removed Discord webhook uploader code and related Discord webhook integration workflows/tests

# ScamScreener 1.3.1 This version keeps the whitelist/review improvements and switches training-data upload to a manual Discord handoff flow. The whitelisted users are ignored by ScamScreener. ## Added - Added whitelist management in-game with both commands and GUI (`/scamscreener whitelist`, `add`, `remove`, `list`, plus `/ss` aliases). - Added a dedicated whitelist config file: `config/scamscreener/scam-screener-whitelist.json`. - Added direct review shortcut commands: `/scamscreener review <playerName>` and `/ss review <playerName>`. - Added a manual upload handoff message that asks to join Discord and includes clickable invite/folder links. ## Changed - Detection now skips whitelist players early in the pipeline (right after mute filtering), before scoring and warnings. - Review auto-marking now follows `AI Auto-Capture` level (`OFF` disables automatic `[S]`; `LOW|MEDIUM|HIGH|CRITICAL` auto-mark only matching-or-higher risk rows). - Whitelist display names are now refreshed toward canonical Mojang names when lookup succeeds. - `/scamscreener upload` now archives training data and prompts manual Discord upload instead of webhook upload. - Upload ToS and README upload/network documentation now reflect player-controlled manual upload behavior. ## Fixed - Fixed review row merge behavior so scored flagged/context rows keep their score signal reliably during review assembly. - Fixed upload status messaging so it no longer implies automatic webhook submission. ## Removed - Removed Discord webhook uploader code and related Discord webhook integration workflows/tests

# ScamScreener 1.2.3 This update improves review accuracy and readability: more decorated chat lines are captured, duplicate self-message echoes are removed, and each review row now shows the mod score. ## Added - Added per-row mod risk score display in the live reviewer, shown directly after `[I] [S] [L]` as `(0)` to `(100)`. ## Changed - Expanded player chat parsing so heavily decorated lines (for example with level/rank tags and extra symbols between them) are recognized more reliably. ## Fixed - Fixed duplicate entries in live review for your own messages by de-duplicating outgoing chat against immediate incoming echo lines. - Fixed missing review captures for some decorated player chat formats that previously failed to parse.

# ScamScreener 1.2.3 This update improves review accuracy and readability: more decorated chat lines are captured, duplicate self-message echoes are removed, and each review row now shows the mod score. ## Added - Added per-row mod risk score display in the live reviewer, shown directly after `[I] [S] [L]` as `(0)` to `(100)`. ## Changed - Expanded player chat parsing so heavily decorated lines (for example with level/rank tags and extra symbols between them) are recognized more reliably. ## Fixed - Fixed duplicate entries in live review for your own messages by de-duplicating outgoing chat against immediate incoming echo lines. - Fixed missing review captures for some decorated player chat formats that previously failed to parse.

# ScamScreener 1.2.1 ScamScreener 1.2.1 mainly expands the in-game review and training-data workflow. ## Highlights ### Added - Added `/ss` as a full alias for `/scamscreener`. - Added direct menu opening on `/scamscreener` and `/ss` (help now via `/scamscreener help` or `/ss help`). - Added `Review Training CSV` in the mod menu next to `Upload Training Data`. - Added full CSV review flow with `Save`, `Save & Upload`, and `Open File`. - Added logged-chat review via `/scamscreener review` and `/ss review` for not-yet-saved chat lines. - Added autocomplete for `/scamscreener review player <playerName>` and `/ss review player <playerName>` with online players that have logged chat entries. - Added automatic training-data size reminders from 500+ entries (initial warning, then every 5 minutes) with clickable upload action. ### Changed - Changed review row state display to native colored `[I]`, `[S]`, and `[L]` markers. - Changed review help routing to `/scamscreener review help` and `/ss review help`. - Changed CSV review editing to row-based handling so duplicate messages can be reviewed independently. ### Fixed - Fixed CSV label save behavior so message quoting (`"..."`) is preserved. - Fixed CSV review ignore behavior so ignored rows are removed on save. ## Compatibility - Minecraft `1.21.10` and `1.21.11`. - Fabric (client-side). ## Note ScamScreener remains client-side. Detection, review, and local training data stay on your client unless you explicitly upload training data.

# ScamScreener 1.2.1 ScamScreener 1.2.1 mainly expands the in-game review and training-data workflow. ## Highlights ### Added - Added `/ss` as a full alias for `/scamscreener`. - Added direct menu opening on `/scamscreener` and `/ss` (help now via `/scamscreener help` or `/ss help`). - Added `Review Training CSV` in the mod menu next to `Upload Training Data`. - Added full CSV review flow with `Save`, `Save & Upload`, and `Open File`. - Added logged-chat review via `/scamscreener review` and `/ss review` for not-yet-saved chat lines. - Added autocomplete for `/scamscreener review player <playerName>` and `/ss review player <playerName>` with online players that have logged chat entries. - Added automatic training-data size reminders from 500+ entries (initial warning, then every 5 minutes) with clickable upload action. ### Changed - Changed review row state display to native colored `[I]`, `[S]`, and `[L]` markers. - Changed review help routing to `/scamscreener review help` and `/ss review help`. - Changed CSV review editing to row-based handling so duplicate messages can be reviewed independently. ### Fixed - Fixed CSV label save behavior so message quoting (`"..."`) is preserved. - Fixed CSV review ignore behavior so ignored rows are removed on save. ## Compatibility - Minecraft `1.21.10` and `1.21.11`. - Fabric (client-side). ## Note ScamScreener remains client-side. Detection, review, and local training data stay on your client unless you explicitly upload training data.

# ScamScreener 1.1.0 ScamScreener 1.1.0 improves in-game review workflows and expands scam education coverage. ## Highlights ### Added - Added review commands: `/scamscreener review manage <alertId>`, `/scamscreener review info <alertId>`, `/scamscreener review player <playerName>`. - Added a dedicated Manage Alert workflow with multi-line labeling (`Ignore -> Scam -> Legit`), save/upload flow, and optional blacklist/block actions. - Added a dedicated Alert Rule Details screen with grouped trigger stages (`Rule`, `Behavior`, `Similarity`, `Trend`, `Funnel`, `AI`) and captured context. - Added expanded education follow-up messages for external redirects, suspicious links, upfront payment requests, account-data requests, fake middleman claims, urgency pressure, trust manipulation, too-good-to-be-true offers, Discord redirects, and funnel patterns. - Added education message control command: `/scamscreener edu disable <messageId>`. - Added Metrics in Menu for some information ### Changed - Changed scam warning actions to centered review tags (`[manage] [info]`) and improved warning layout alignment. - Changed review/info context assembly to merge recent captured lines with evaluated pipeline lines (normalized and deduplicated). - Changed default alert-threshold migration behavior to enforce a one-time minimum of `MEDIUM` for older unmigrated configs. ### Removed - Removed deprecated training-capture commands `/1`, `/0`, `/scamscreener ai capture`, and `/scamscreener ai capturebulk`. ### Fixes - Fixed alert info context so scored pipeline messages are consistently shown in captured-message context. ## Compatibility - Minecraft `1.21.11` and `1.21.10`. ## Note ScamScreener is client-side. Detection, scoring, and local review/training data stay on your client unless you explicitly upload training data.

# ScamScreener 1.1.0 ScamScreener 1.1.0 improves in-game review workflows and expands scam education coverage. ## Highlights ### Added - Added review commands: `/scamscreener review manage <alertId>`, `/scamscreener review info <alertId>`, `/scamscreener review player <playerName>`. - Added a dedicated Manage Alert workflow with multi-line labeling (`Ignore -> Scam -> Legit`), save/upload flow, and optional blacklist/block actions. - Added a dedicated Alert Rule Details screen with grouped trigger stages (`Rule`, `Behavior`, `Similarity`, `Trend`, `Funnel`, `AI`) and captured context. - Added expanded education follow-up messages for external redirects, suspicious links, upfront payment requests, account-data requests, fake middleman claims, urgency pressure, trust manipulation, too-good-to-be-true offers, Discord redirects, and funnel patterns. - Added education message control command: `/scamscreener edu disable <messageId>`. - Added Metrics in Menu for some information ### Changed - Changed scam warning actions to centered review tags (`[manage] [info]`) and improved warning layout alignment. - Changed review/info context assembly to merge recent captured lines with evaluated pipeline lines (normalized and deduplicated). - Changed default alert-threshold migration behavior to enforce a one-time minimum of `MEDIUM` for older unmigrated configs. ### Removed - Removed deprecated training-capture commands `/1`, `/0`, `/scamscreener ai capture`, and `/scamscreener ai capturebulk`. ### Fixes - Fixed alert info context so scored pipeline messages are consistently shown in captured-message context. ## Compatibility - Minecraft `1.21.11` and `1.21.10`. ## Note ScamScreener is client-side. Detection, scoring, and local review/training data stay on your client unless you explicitly upload training data.

# ScamScreener 1.0.1 ScamScreener 1.0.1 improves the training-data upload flow and keeps the command/menu behavior consistent. ## Highlights In 1.0.1 - Full support for Minecraft `1.21.11` and `1.21.10`. - `/scamscreener ai train` was replaced by `/scamscreener upload`. - Upload now requires the active file `config/scamscreener/scam-screener-training-data.csv`. - Upload no longer falls back to `config/scamscreener/old/training-data/*.old.*`. - Upload is gated behind a one-time ToS consent screen. - Training capture now also includes outgoing chat command text (`/msg`, `/r`, party, guild, `ac`). ## What You Get As A Player - Live scam warnings in chat with risk levels and clear trigger details. - Local blacklist warnings in relevant contexts (for example trade, party, and co-op). - Quick action buttons in warnings (`legit`, `scam`, `blacklist`, `block`). - Optional auto-leave from parties when a blacklist hit is detected. - Custom mute filter (including regex) to hide scam spam immediately. - Outgoing safety guard for risky content (email, Discord invites, `/coopadd`) with controlled bypass. - In-game settings GUI (`/scamscreener settings`) for fast configuration. - Local AI model with update workflow (download/accept/merge/ignore), including SHA-256 integrity checks. ## Quick Commands - `/scamscreener settings` Open Settings Menu - `/scamscreener add <player> [score] [reason]` Add Player to Blacklist - `/scamscreener ai update` Check for Model Update - `/scamscreener upload` Upload your Training Data - `/scamscreener mute <pattern>` Mute messages with patterns ## Note ScamScreener is client-side. Core detection, scoring, and local data stay on your client.

# ScamScreener 1.0.1 ScamScreener 1.0.1 improves the training-data upload flow and keeps the command/menu behavior consistent. ## Highlights In 1.0.1 - Full support for Minecraft `1.21.11` and `1.21.10`. - `/scamscreener ai train` was replaced by `/scamscreener upload`. - Upload now requires the active file `config/scamscreener/scam-screener-training-data.csv`. - Upload no longer falls back to `config/scamscreener/old/training-data/*.old.*`. - Upload is gated behind a one-time ToS consent screen. - Training capture now also includes outgoing chat command text (`/msg`, `/r`, party, guild, `ac`). ## What You Get As A Player - Live scam warnings in chat with risk levels and clear trigger details. - Local blacklist warnings in relevant contexts (for example trade, party, and co-op). - Quick action buttons in warnings (`legit`, `scam`, `blacklist`, `block`). - Optional auto-leave from parties when a blacklist hit is detected. - Custom mute filter (including regex) to hide scam spam immediately. - Outgoing safety guard for risky content (email, Discord invites, `/coopadd`) with controlled bypass. - In-game settings GUI (`/scamscreener settings`) for fast configuration. - Local AI model with update workflow (download/accept/merge/ignore), including SHA-256 integrity checks. ## Quick Commands - `/scamscreener settings` Open Settings Menu - `/scamscreener add <player> [score] [reason]` Add Player to Blacklist - `/scamscreener ai update` Check for Model Update - `/scamscreener upload` Upload your Training Data - `/scamscreener mute <pattern>` Mute messages with patterns ## Note ScamScreener is client-side. Core detection, scoring, and local data stay on your client.

# ScamScreener 1.0.0 ScamScreener 1.0.0 is the first major stable release for Hypixel SkyBlock players. It helps you detect scam patterns in chat early and react quickly. ## What You Get As A Player - Live scam warnings in chat with risk levels and clear trigger details. - Local blacklist warnings in relevant contexts (for example trade, party, and co-op). - Quick action buttons in warnings (`legit`, `scam`, `blacklist`, `block`). - Optional auto-leave from parties when a blacklist hit is detected. - Custom mute filter (including regex) to hide scam spam immediately. - Outgoing safety guard for risky content (email, Discord invites, `/coopadd`) with controlled bypass. - In-game settings GUI (`/scamscreener settings`) for fast configuration. - Local AI model with update workflow (download/accept/merge/ignore), including SHA-256 integrity checks. ## New In 1.0.0 - Full support for Minecraft `1.21.11` and `1.21.10`. - New community training flow: `/scamscreener ai train` archives your local training data and guides you to upload it for community model training. - New `Upload Training Data` button in the settings menu. - Upload guidance now includes a clickable local folder path and clickable SkyblockEnhanced Discord link. - Blacklist warning header was fixed (`BLACKLIST WARNING`). ## Quick Commands - `/scamscreener settings` - `/scamscreener add <player> [score] [reason]` - `/scamscreener ai update` - `/scamscreener ai train` - `/scamscreener mute <pattern>` ## Note ScamScreener is client-side. Core detection, scoring, and local data stay on your client.

# ScamScreener 1.0.0 ScamScreener 1.0.0 is the first major stable release for Hypixel SkyBlock players. It helps you detect scam patterns in chat early and react quickly. ## What You Get As A Player - Live scam warnings in chat with risk levels and clear trigger details. - Local blacklist warnings in relevant contexts (for example trade, party, and co-op). - Quick action buttons in warnings (`legit`, `scam`, `blacklist`, `block`). - Optional auto-leave from parties when a blacklist hit is detected. - Custom mute filter (including regex) to hide scam spam immediately. - Outgoing safety guard for risky content (email, Discord invites, `/coopadd`) with controlled bypass. - In-game settings GUI (`/scamscreener settings`) for fast configuration. - Local AI model with update workflow (download/accept/merge/ignore), including SHA-256 integrity checks. ## New In 1.0.0 - Full support for Minecraft `1.21.11` and `1.21.10`. - New community training flow: `/scamscreener ai train` archives your local training data and guides you to upload it for community model training. - New `Upload Training Data` button in the settings menu. - Upload guidance now includes a clickable local folder path and clickable SkyblockEnhanced Discord link. - Blacklist warning header was fixed (`BLACKLIST WARNING`). ## Quick Commands - `/scamscreener settings` - `/scamscreener add <player> [score] [reason]` - `/scamscreener ai update` - `/scamscreener ai train` - `/scamscreener mute <pattern>` ## Note ScamScreener is client-side. Core detection, scoring, and local data stay on your client.

### Added - Added Stonecutter setup to support Minecraft `1.21.11`.

### Added - Added Stonecutter setup to support Minecraft `1.21.11`. **No Changes for MC 1.21.10**

### Added - Added funnel coverage for an upfront-payment variant (`OFFER -> PAYMENT`) with automated tests. ### Changed - Local AI trainer now preserves the current model version when retraining instead of forcing a static version value. ### Fixed - Model updater now treats remote models with equal or older model version as up-to-date and skips unnecessary update prompts. - Funnel detection and AI funnel context now also handle the upfront-payment funnel path (`OFFER -> PAYMENT`).

### Fixed - Fixed false `sha256 mismatch` errors when downloading AI model updates by tolerating equivalent UTF-8 payload variants (LF/CRLF, BOM/no BOM). - Fixed model release hash generation so `scripts/model-version.json` SHA-256 matches GitHub raw model payload bytes.

### Added - New `FunnelSignalStage` with per-player sequence tracking (ring buffer + TTL) and funnel evidence in warnings. - Intent tagging layer (`IntentTagger`) with tags for offer/rep/redirect/instruction/payment/community anchor. - Obfuscation-aware redirect detection for variants like spaced or leetspeak platform mentions. - Configurable funnel settings and intent patterns in rules config (window size/time, weights, negative-intent patterns). - Channel-aware message context parsing (`pm`, `party`, `team`, `public`) for pipeline and training features. - New AI feature stack with dense feature space + funnel context tracker (`AiFeatureSpace`, `AiFunnelContextTracker`). - Expanded training CSV schema with conversation/window and funnel-aware fields (`window_id`, intent fields, stage hits, funnel metrics, hard negatives, sample weights). - Additional AI rule signal: `LOCAL_AI_FUNNEL_SIGNAL` plus dedicated funnel tuning config (`localAiFunnelMaxScore`, `localAiFunnelThresholdBonus`). - Warning action tag `[block]` to run `/block <player>` from risk messages. - Model update metadata now includes model schema version in `scripts/model-version.json`. ### Changed - Detection pipeline order updated so Funnel runs before AI scoring, allowing the model to consume live funnel context. - Local AI model format and training were reworked to support dense+token scoring and a dedicated funnel head (`funnelHead`) in the same model file. - Local AI trainer now trains main and funnel heads from the same CSV (funnel labels inferred from funnel markers). - Auto-captured training samples now persist full detection-aware feature rows instead of plain text-only rows. - Local AI scoring can emit multiple AI signals in one pass (general + funnel-focused). - Funnel-AI trigger threshold and max score are sourced from `ScamRulesConfig` (no hardcoded trigger values). - Model updater validates schema and preserves/normalizes funnel-head weights on apply/merge. - Model version update script validates schema before calculating SHA and publishing metadata. - Warning action tags use direct one-click command execution (`[legit]`, `[scam]`, `[blacklist]`, `[block]`). - Internal player tracking/training/debug flow now uses anonymized speaker keys and anonymized training capture context. ### Removed - Removed legacy training flag pipeline (`TrainingFlags`, `TrainingTypeAiClassifier`). - Removed legacy local-model migration path from root config for AI model loading. ### Fixed - System-message filtering was tightened for party/trade/co-op lines so they stay out of detection/training. - Chat lines starting with `[NPC]` are treated as system messages and excluded from player detection/training capture. ### Kept - Risk warning messages still show the real player name for in-game risk message context.

### Added - AI updater join-notify controls via `/scamscreener ai update notify [on|off]` (no args = status) and the AI update menu toggle (`Join Up-to-Date Message`). - Coop blacklist warning coverage expanded for co-op join request, `You invited <name> to your co-op!`, and `<name> joined your SkyBlock Co-op!`. - Coop safety flow for `/coopadd <player>` with `[BYPASS]` confirmation. - Location backbone service for scoreboard-based island detection. - `SkyblockIsland` enum with island names and alias matching. - Thanks to Kd_Gaming1: ModMenu integration as soft dependency with config screen support from mod list. ### Changed - `autoleave` command behavior updated: no args now shows status; explicit toggle remains `on|off`. - `ai update notify` behavior updated: no args now shows status; explicit toggle remains `on|off`. - `/coopadd` is now always blocked first and requires explicit `[BYPASS]`. - Mod contact metadata now defines an issues URL so the ModMenu Issues button opens https://github.com/Tangos-Mods/ScamScreener/issues/new. ### Fixed - Blacklist warning lookup now falls back to name-based entries when UUID lookup is unavailable. - Coop blacklist bypass message formatting now uses the shared `MessageBuilder` style. - ModMenu config-screen implementation fixed (correct `Screen` return type in the config factory). **As in previous Versions: Please provide scam messages in the issues Tracker on GitHub so I can train the model**

## Added - New GUI system exposed via `/scamscreener settings`. - Dedicated submenus for `Rules`, `Debug`, `Messages`, `Blacklist`, and `AI Update`. - AI update menu actions: `Accept`, `Merge`, `Ignore`, plus `Force Check`. --- This is still a beta. Please report any issues or feature requests to GitHub