Prototype Mentat/Bureaucrat score effects, automatic Sandmaster board spice, and Rihani inspection plus committed exchange. Preserve private views, physical custody, saved winner cleanup and captive/Suk sequencing. Remaining normal bands, missing skill families and combined modules remain gated. Validation: 45 focused tests; typecheck, lint, 4465 offline tests, production build and 43 HTTP tests passed. Independent rules/privacy/persistence review completed. New isolated browser setup exercised inspection, draw, pending refresh, exchange and completed refresh. Two genuine four-profile Basic/Advanced samples completed without rejected candidates but did not trigger Rihani. Preserved all 393 opening rooms and original seat/recovery rows (427 rooms after QA/HTTP). SQLite integrity and foreign keys passed. Server remained healthy after the reported outage; no restart or reset. Verified source tree: c64ce9537269de94c0d785bc08b17346e826c5797352a636771e1dc6ce5e4859
87 lines
5.4 KiB
Markdown
87 lines
5.4 KiB
Markdown
# Leader Skill battle effects
|
|
|
|
13 September 2026. **Prototyped, with Partial rules coverage.** This follow-up
|
|
connects both Rihani Decipherer bands and the lower battle bands of Mentat,
|
|
Bureaucrat and Sandmaster. Normal Mentat questioning, Bureaucrat payments and
|
|
Sandmaster movement collection remain missing, as do Spice Banker, Diplomat and
|
|
Smuggler. Public module starts and publication remain gated.
|
|
|
|
## Source contract
|
|
|
|
The [GF9 CHOAM & Richese rulebook, page 9](https://www.gf9games.com/dune/wp-content/uploads/2021/11/CHOAM-Rulebook-low-res.pdf)
|
|
provides the common skill procedure. The printed cards and publisher designer
|
|
walkthrough supply the individual effects; see the [component inventory](LEADER_SKILLS_RULES.md).
|
|
|
|
- [Mentat, 17:37](https://www.youtube.com/watch?v=XT_azRVLq_0&t=1057s): the
|
|
surviving selected skilled disc gains two battle strength. Its printed value,
|
|
bounty and Zoal source value stay unchanged.
|
|
- [Bureaucrat, 15:01](https://www.youtube.com/watch?v=XT_azRVLq_0&t=901s): the
|
|
surviving selected skilled disc subtracts one from its opponent's total per
|
|
distinct stronghold that opponent occupies. Contested and storm-covered
|
|
occupation count; advisors and virtual victory control do not. There is no
|
|
implicit zero clamp. The shared board helper recognizes HMS and Jacurutu,
|
|
though combined-module play remains gated.
|
|
- [Sandmaster, 15:48](https://www.youtube.com/watch?v=XT_azRVLq_0&t=948s): a
|
|
surviving skilled winner adds three board spice to an existing pile in the
|
|
battle territory automatically. This grants no direct income and does not
|
|
require surviving forces. No pile means no addition. Multiple positive piles
|
|
are guarded before the final battle vote pending a sector-placement ruling.
|
|
- [Rihani, 12:49](https://www.youtube.com/watch?v=XT_azRVLq_0&t=769s): a face-up
|
|
native trainer grants the normal private inspection after a victory using
|
|
another leader. A surviving selected native skilled winner receives both
|
|
bands. A surviving captured skilled winner receives only the lower band.
|
|
|
|
Rihani first privately inspects two random physical Traitor Deck cards and
|
|
reshuffles them. It then offers the separate optional two-card draw. Choosing
|
|
to draw commits the player to keep one newly drawn card, publicly reveal and
|
|
return one unused old card, and secretly return the unkept new card. Those two
|
|
returns are shuffled into the actual deck. There is no post-draw decline or
|
|
return-both-new alternative. The two samples can contain the same identity.
|
|
|
|
**Ordering inference:** normal-before-lower follows the printed order and
|
|
designer narration; no explicit priority sentence was found. **Unused** means
|
|
not actually called, rather than merely known or publicly disclosed. Returning
|
|
an old card announces its name without adding it to the actual-use register.
|
|
A Traitor Card used to win this battle is already excluded from the exchange.
|
|
No eligible old card means no optional draw. There is no partial draw fallback
|
|
when fewer than two physical deck cards are available.
|
|
|
|
The [Ixian/Tleilaxu rules](https://www.gf9games.com/dunegame/wp-content/uploads/2020/09/IxianAndTleilaxuRulebook.pdf)
|
|
place skill resolution before faction capture and Face Dance. Tleilaxu Rihani
|
|
must instead exchange an unrevealed Face Dancer. That combination, foreign
|
|
gholas and the skill lifecycle after non-kill Face Dance remain unimplemented.
|
|
The connected entry supports base factions in Basic/Advanced with Leader Skills
|
|
alone. These boundaries are implementation limits, not prohibitions in the game.
|
|
|
|
## Connected behavior and checks
|
|
|
|
Shared battle quotes calculate score changes; bots use the same public
|
|
stronghold count. Sandmaster resolves automatically. Rihani offers owned
|
|
draw/decline controls, then separate inspectable keep-new and reveal-old
|
|
selections. All four AI profiles use only their permitted view. Only the owner
|
|
receives inspection history and drawn identities. The public log names the
|
|
returned old card, never the kept or unkept new identity.
|
|
|
|
Physical snapshots, an event-bound obligation and saved stages preserve Rihani
|
|
through casualties, winner cleanup and captive return. Independent review found
|
|
that completed receipts were not bound to custody during cleanup; this now
|
|
rejects changed deck order after either decline or exchange. JSON and SQLite
|
|
restart tests cover both choices, concurrent writes, stale replays, normal versus
|
|
captured eligibility, actual-use exclusion and native Suk rescue preceding a
|
|
captured Rihani exchange. Focused checks also cover score-boundary winners,
|
|
dead/bluffed skills, Sandmaster without surviving forces and UI privacy.
|
|
|
|
Fresh browser room `HNGSR8GB` used genuine module setup, then a separately
|
|
staged conserved battle. The human inspected two private cards, drew two
|
|
separate cards, refreshed the pending exchange and selected a keep and public
|
|
return. Completion and refresh restored the kept Traitor and winner card cleanup.
|
|
This is targeted browser evidence, not a naturally reached complete game.
|
|
Two genuine four-profile Basic/Advanced samples completed in 92/188 actions,
|
|
with 3/6 JSON continuations and no rejected candidates; neither reached a Rihani
|
|
effect. They establish sample continuation, not comprehensive skill acceptance.
|
|
|
|
The running server responded after the reported outage. A fresh online backup
|
|
captured all 393 opening rooms. No restart or reset was needed. Required broad
|
|
checks and final saved-game preservation are recorded in the checkpoint commit
|
|
and private source-bound report.
|