Files
Dune/docs/SUK_GRADUATE_RULES.md
T
Quintine af360c836b Connect Suk Graduate casualty rescues and saved continuation
Implement both rescue bands with legal physical casualty routing, owner controls,
all four bot profiles and durable battle receipts. Preserve paid support, elite
identity, killed/returned leader custody and replay rejection. New Advanced
Atreides skill choices exclude Suk while the KH loss-count ruling is pending;
legacy assigned combinations retain their preflight guard. Public gates remain.

Independent review fixed stale dead trainer assignments and detached rescue
receipts. Final focused Suk/Leader Skills union: 46 passed. Typecheck, lint,
4,452 offline tests, production build and 43 HTTP tests passed on unchanged
source tree 3b76239a46d0ba874299495e64295a5eaaa8f2415c3e9a041bee944bbce50c5a.

Actual browser rescue and refresh retained counters, private hand and pending
winner card cleanup. Two four-profile Basic/Advanced smoke games completed
with JSON restoration and no rejected actions; neither exercised a Suk rescue.
After the outage report the server was healthy. All 360 pre-check rooms and
original seat/recovery records are unchanged; checks added 33 rooms. SQLite
integrity and foreign-key checks pass. No server restart or database reset.
2026-09-13 19:02:32 +10:00

6.1 KiB
Raw Blame History

Suk Graduate rescue prototype

13 September 2026. Prototyped / Partial. This connects both Suk Graduate bands to real battle resolution, player choices, AI and saved continuation. Complete module, faction and publication gates remain closed.

Source contract

Use the physical card and the Leader Skills source archive. The GF9 CHOAM & Richese rulebook page 9 governs visibility, survival and skill-before-faction ordering; its page 12 Q&A gives a captor only the lower band of an actually used captive. The designer walkthrough at 11:10–11:59 expressly distinguishes the normal and selected-leader rescues and says they do not accumulate. The archived card photograph supplies the precise quantities.

After an ordinary win with physical casualties, a face-up native Suk Graduate returns exactly one chosen casualty to reserves instead of the Tanks. This band is mandatory. When the actual skilled leader is used and survives, the player may instead save zero through three casualties, limited by the raw casualties. If any are saved, exactly one stays in the battle territory and the rest return to reserves. Declining this lower band does not trigger the normal band.

A concealed skill with another selected leader grants neither band. A killed selected skilled leader grants no rescue. A different selected leader may die while the face-up native trainer remains alive and grants the normal rescue. A traitor victory loses no winning forces; mutual traitors and explosions have no winner. None creates a Suk rescue. The zero-dial case also creates no window.

Physical settlement and continuation

The existing legal raw normal/elite casualty choices are retained. Choose that allocation first; Suk then redirects only those actual counters. Paid support is unchanged. Saved elites remain elites in reserves or on the board. The normal band resolves automatically when only one counter type can be saved; the lower band's optional choice remains with its owner.

The receipt freezes revealed eligibility before the live battle is cleared or a used captive returns. It records the battle, raw commitment, source groups, physical pool and remaining card cleanup. A separate obligation in the last battle context prevents a missing receipt from silently bypassing rescue. Actions, views and automatic continuation validate it. A stale event, altered allocation, changed physical pool or missing choice rejects without mutation. The server alone projects the legal rescue options; the receipt is not exposed.

The current engine assigns a raw casualty mix across sectors in existing location order. A saved counter that remains stays in that original sector; there is no free sector relocation. This is an explicit implementation inference: the card specifies a territory, not how to select sectors. Saved counter types and any eligible sector choice remain visible to the player.

Suk settles before winner card choices and faction aftermath, including capture and Face Dance. Mandatory loser-card cleanup uses its existing continuation. The exact ordering against future combined optional card interruptions still needs integrated acceptance. Homeworld custody, Nexus and expansion-faction combinations are explicitly rejected before resources or traitor votes commit; the genuine Leader Skills initializer currently admits base factions only.

Atreides loss-count question

The base Advanced rules activate Kwisatz Haderach after seven force losses. Suk describes rescued counters as losses redirected away from the Tanks, and the designer describes the retained counter as having been killed. Counting the raw dialed casualties is therefore the stronger textual inference, but the retrieved publisher/designer sources do not expressly settle this combination. The user question is pending. New Advanced Atreides setup and revival offers retain their physical cards but make Suk unavailable, explain why and allow the other offered card. The server, private view, controls and bots share this prototype restriction. Already-assigned saved combinations retain the battle preflight guard before final traitor votes or resources commit. This availability restriction is not an official rule. Other rescued counters are kept out of the physical Tank and ordinary battle-loss totals in this bounded prototype.

Evidence and remaining work

Focused regressions cover genuine module setup followed by conserved targeted battles; both bands, no stacking, decline, killed leaders, captured-leader return, normal/elite routing, unchanged support and all four bot profiles. JSON and SQLite store restart retain the pending choice. Concurrent submissions commit one rescue; the stale submission changes no counters. Component checks cover clear destinations, required selection and disabled controls.

Fresh browser room BD33WTTV used genuine Emperor/Guild module setup and selected Suk Graduate/Hasimir Fenring. A separately staged, conserved battle position used one physical Baliset from the deck. Actual browser rescue saved three ordinary casualties: one stayed in sector 10 and two returned to reserves. The result was two board forces, seventeen reserves and one Tank. Refresh after another reported outage retained that result, the private Crysknife/Baliset/Burseg and winner card cleanup. This was a targeted battle, not a naturally reached full game.

Four-profile samples completed Basic seed 20260918 (92 actions, three JSON restorations) and Advanced seed 20260919 (188 actions, six restorations), with no rejected actions and conserved cards/forces after each action. Neither sample encountered an eligible Suk rescue; effect evidence comes from the targeted tests and browser journey. Checkpoint-wide results and final preservation are recorded in the commit and source-bound private report.

These tests do not certify the full expansion, every combination, AI strength ordering or complete human games. Seven other skill effects remain missing; retain the full scope and pending rulings.