Dune/docs/STRONGHOLD_CARDS.md

Advanced Stronghold Cards

Historical development module checkpoint, 7 September 2026: the six printed effects, public card custody, mobile declaration, support accounting, AI choices and readable faces are integrated. The later user-authorized classic Advanced preview already permits Stronghold Cards; preserve that access. Complete Advanced/expansion release and arbitrary combinations remain gated.

Independent Treachery composition — 16 October 2026

The fresh independent-card entry retains original unused Advanced Strongholds at2–6 seats and optional Tech3–6, with clean classic/selected Ix/CHOAM native/base rosters. Classic and selected native Homeworlds compose the same original phase modules. Actual end-Mentat earns ownership; no card is assigned by a fixture. Arrakeen funds actual support (bank2/personal1 for support3); native Ix earns Tabr before a real next-turn attack and mixed Reinforcements fee/Harass return. Original token rewards/held effects stay. Stone/Mirror uses its separate selected-family contract. Other overlays, public starts and save conversion remain gated. See the shared phase contract and World composition.

Semuta composition — 11 October 2026

The fresh semuta entry with native Richese, CHOAM and optional Ix decks now preserves original unused Strongholds in Advanced2–6; optional unused Tech requires3–6. It reuses the original native setup/roster prerequisites and does not modify physical card quantities, effects or public release gates.

Closing CHOAM Gamont may offer its used Treachery card to Semuta before the original owner finishes. Stronghold custody stays unclaimed across that reaction and is assigned only by actual final Mentat using the changed board. The six module cards are not Treachery hand/discard cards or Semuta targets. No Skills, Homeworlds, Nexus, Discovery, Basic Strongholds, E3 mixture or played-game conversion is added by this entry.

Nine actual fresh Advanced setup/first-Storm cases and twelve closing runtime cases cover both configurations, claim/pass and four legal profiles. Both 390px styled table paths retain the original closing control, reach turn3 Storm and display Arrakeen held by Atreides/Carthag unclaimed at claim2 after Gamont removed the sole Carthag force. This is bounded local evidence, not complete Stronghold/Semuta/mode or deployed acceptance.

Original Discovery composition — 5 October 2026

Fresh original discovery entry preserves selected unused Strongholds6 in Advanced2–6 classic/supported E1/E2/standalone E3, with optional Tech3–6. The separate classic leader-skills --discoveries entry preserves base33/all14/Discovery7/8 and the same original Advanced Strongholds/Tech. The supplied authorized source physical21 permits variants together; this bounded entry does not open every unresolved interaction or public mode.

Original first end-Mentat ownership is required. Benefits stay in the held card's named territory or the actual HMS copy; Cistern/Shrine and other revealed nested sites do not inherit another site's defense or support. Jacurutu adds a stronghold victory location, not a seventh physical card. Free entry, paid nested forces/markers, native typed casualties, trained rescue and original cleanup/Tech/Face Dance retain their existing source identities. No held receipt or paid result is injected to grant an advantage.

Basic Strongholds, native Discovery skill families, Homeworlds, independent variants/previews, pending rules, public starts and save conversion remain separate. Existing Nexus/Discovery/Strongholds is unchanged. The Discovery contract records the original CLI and genuine-runner entries and bounded evidence.

Affected77 and post-repair coalition39, types/lint/build and actual programmes pass. Phone UWEH7WNBv14 retains natural Smuggler/held Arrakeen/four paid/free Cistern counters. Native4LBU44PTv21 records human HMS copy/support3/bank2/ own1, actual CHOAM income1, three physical Cyborg losses, protected Shrine4/ two Cyborgs and original Production winner transfer. The five completed genuine six-seat samples include one actual mandatory Ecaz repair/resume; the sixth Richese sample remains at the existing exhausted-cache ruling. Counts, source trees, conserved staging and390px refresh are recorded in the linked contract; these proofs do not certify complete modules or deployment.

Native Tech composition without Skills — 4 October 2026

Fresh Advanced stronghold-factions may preserve original unused Tech Tokens at3..6 seats, without Leader Skills, in selected E1/E2 native families or standalone Ecaz OR Moritani plus classic opponents. Original selected decks, cache, native first-Storm Tech owners and actual end-Mentat Stronghold claims remain. Tech-only Basic/Advanced uses existing factions entry; read the native composition contract.

The actual selected Ecaz lead receives its own held-card effect and original winner Tech; neither card advantages nor split token holdings are shared. Original typed casualties/payments, losing Tuek income, Moritani post-loss assassination and original winner rewards before Face Dance remain distinct. Two-player Tech, Basic Strongholds, mixed E3, other overlays and public starts stay gated; existing unresolved rulings are unchanged.

New32/affected187 cases, types/lint/build, twelve actual native programs and all four legal Face Dance policies pass. Fifteen genuine3-seat module games complete4,902 actions/no rejection/125 JSON continuations. Original human BZVLA22D v21/Mentat retains actual selected-lead Tabr income and two original Tech Tokens; YENG8B6U v15/Mentat retains losing Tuek income, printed Master3 assassination, original winner Tech and one private replacement. Read source-bound evidence.

Standalone E3 skills and optional Tech — 4 October 2026

Fresh original native Ecaz OR Moritani ecaz33/all14 skill profiles additionally retain unused Stronghold Cards in Advanced2..6, with original Tech3..6. Actual end-Mentat/local held-card effects and native setup/wallets remain. Ecaz five-disc/temporary Duke and Advanced-Hark exclusion, plus original Moritani skill-first assassination/normal-call forfeiture/replacement remain. E3 pairs, mixtures, allied Occupy skills and other overlays stay excluded. Human390px 334W75B6 v13/Collection combines Carthag Shield/normal Warmaster1/acquired Duke6: real7–6 poison survival, living release, native training/cards and original Tech. AL6WCG8B v13/Mentat retains Guild Suk before unused-disc death, losing Tuek4/support1, printed3 and one replacement. New15/affected152, types/lint/build, nine programs and eight E3 natural games pass (1,531 actions/no rejection,38 JSON continuations). Read the existing combined contract. Full-module/public/deployed acceptance remains open.

Advanced Leader Skills composition — 4 October 2026

Fresh local leader-skills, now including mandatory normal Banker income, preserves unused six Stronghold Cards with all14 skills in Advanced classic or supported E1/E2 native2..6-seat rosters. Optional original Tech requires3..6; native wallets, decks/cache, setup and actual end-Mentat claims remain. The supplied authorized Advanced source physical21/printed19 permits variants together. Basic Strongholds, E3/other modules, public starts and save conversion stay excluded.

Original support subsidy changes actual payer debit, not rescue losses or the normal Banker threshold. Native CHOAM still earns on bank-funded support. Physical Carthag/normal discipline, holder-only Habbanya tie, Tabr dial income, Tuek played-Worthless income and HMS pre-plan copy compose source-clear skills. Suk and original native cleanup finish before optional Tech and Face Dance; the original winner retains earnings/cards, and killed training returns once. Copied Diplomat/modified defense and active skill/Stone interpretations remain guarded. No card custody is assigned to create a case.

New21, affected277 and final shared fixture-consumer72 cases, types/lint/build, twelve actual programs and twenty-two genuine paired/triple games pass (6,672 accepted actions/no rejection,171 JSON continuations). Human390px HERCFX3B v20/Mentat pays actual support4 after bank2, rescues3, keeps Shield and original two Tech while separate Banker fronts1 then collects 12→13 at actual Mentat. PG2GR3RC v26/Collection declares real HMS copy, bank-funded1/own0, typed4 Suboid+2 cyborg loss/rescue before sole Axlotl reward. All original responses, market finish and phase-opening readiness remain. Read the combined scope and exact outcomes. This is bounded Development Verified, not complete-module/public/deployed proof.

Standalone E3 Stronghold composition — 4 October 2026

Fresh stronghold-factions additionally admits Advanced native Ecaz OR native Moritani with classic opponents and the exact ecaz Treachery deck, all six Stronghold Cards and no other overlays. Moritani excludes Harkonnen under its original assassination/Duke boundary. A native Ecaz/Moritani pair and E1/E2 mixtures remain separate; no Basic Stronghold start, public expansion start or played-game conversion is added. Original faction setup and first actual end-Mentat ownership remain.

Ecaz's reciprocal co-occupation selects Ecaz as card holder. The chosen battle lead gets only its own held advantage: Ecaz-led Arrakeen can contribute up to2 bank spice to actual ally-variable support, while fixed Ecaz strength is free. An ally-led plan cannot borrow Ecaz's card. Original Habbanya tie priority and Tabr opposing-dial income also belong only to the selected holder; physical fixed/variable casualties and later shared-city Collection remain separate.

Moritani's losing Tuek holder receives its printed Worthless income without winning. Original post-loss assassination follows real support/card income, retains printed disc bounty, retires one physical Traitor and draws one private replacement at actual Mentat. Winner cleanup and dial losses still finish once; a normal Traitor revelation forfeits assassination for the game. Original Terror placement remains free. Carthag Shield-as-Snooper follows its actual holder, not the winning faction or another participant.

Focused20/20 and types/lint pass. Parent actual controlled programs observe Ecaz's native wallet1 paying1 of support3 with bank2, selected Guild's native wallet5 paying all3 with bank0, followed by separate Collection2. Both keep Ecaz2/Guild1 physical survivors. Native Moritani wallet12 pays support1, earns Tuek4 and printed assassination3, reaching18 at actual Mentat with one replacement and no repeated payment. These controlled cases follow real setup, alliances and card claims, not claimed natural battle histories.

The affected native/CLI union passes179/179. Final Richese cases8/8, types, lint and build pass after choosing legal first Storm dials that keep the controlled Richese battlefield outside the next original random Storm band. Obsolete source-inventory and incidental starting-wallet/placement pins were removed, not updated to new incidental values.

Four genuine E3 two/three-seat games complete642 accepted actions, no rejections,15 JSON continuations and3 battles. These samples contain no natural assassination; the assassination evidence is the controlled original program and the human continuation. Original CLI XK8X548M (Ecaz) and HALWWQ6V (Moritani), both v6, remain at genuine Traitor setup with all six Stronghold Cards unowned. Human390px N5SQHZE4 chooses Ecaz lead and seals total6/support3/Sanya4/Shield. The editor shows fixed Ecaz3 plus Guild-variable3, bank2/personal1. Original resolution records native wallet1→0 at winner cleanup, Ecaz3/Guild3 casualties, then separate city Collection2 each; refresh reaches v18/Mentat with Ecaz2/ Guild1 survivors, Ecaz wallet2 and Guild wallet7. Shield remains physically held. Human390px ANUJE8L5 starts at the original post-loss opportunity with wallet15 after support1/Tuek4. Actual Master Bewt reveal earns printed3, wallet18; Guild winner cleanup applies its three dial casualties once. Original Collection and actual Mentat reach v11 with one private Burseg replacement and the declared Master Bewt card face up. Both enlarged components are readable at390px. Conserved staging is explicit; no natural-history, full-combination, public-start or deployed acceptance claim is made. Final guide10/10, types/lint/build and the rebuilt390px guide pass (351px topic, no page overflow). The final isolated worker reload retains human v18/v11 at actual Mentat; owned temporary drivers and managed tabs are removed.

Expanded E1/E2 faction rules

The same fresh stronghold-factions profile now admits selected Ix/Tleilaxu and CHOAM/Richese native families with classic opponents, Advanced only, two through six seats and Stronghold Cards alone. At least one native expansion faction and its selected distinct ix/choam decks are required. CHOAM-only ordinary deck35; Ix or both47; native Richese cache10 stays separate. Standalone E3 composition follows the separate section above; E1/E2/E3 mixtures and other overlays remain excluded. These rule paths are Partial / Development Verified, not complete faction/module acceptance.

Native No-Field0/3/5 control/reveal and real cache-acquired Stone outcomes, plus native Face Dancer zero/partial/army replacement and next-turn card control are integrated. Targeted16/16 rule cases, types/lint and11 actual rule-program cases pass. Eighteen unstaged2..6-seat paired/mixed games complete:11,066 accepted actions, zero rejections,290JSON restores. An old Ixian/CHOAM-only report header was corrected to list per-family decks/cache; an actual three-family report smoke completes with the corrected descriptor.

Human390px controls used original CLI setups: G7VDY6KB v12 chose Stone kill mode and completed the Habbanya-winning battle with real Stone retained; 736HJUD3 v8 selected two reserves for Face Dance and replaced the original winner while its Stronghold card remained with Ixians until the turn boundary. No new save/privacy/custody audit lane, full-suite or historical-seat campaign. Mixed No-Field/full-plan and unresolved Stone/Poison Tooth cases stay guarded. Final coherent rule/guide26/26, check:quick, build and actual expanded390px guide pass;205 changed document links are valid. Owned temporary drivers and the test tab/service are removed/closed.

Native Ixian and CHOAM integration — 2 October 2026

Fresh local stronghold-factions integration has Partial / Development Verified rule-path coverage: Advanced only, native Ixians and CHOAM plus unique classic opponents, two through six seats, exactly Ix and CHOAM Treachery decks and Stronghold Cards. The combined deck has 47, not49, physical cards: CHOAM replaces the two shared Ix cards rather than adding duplicate identities. No other faction or optional-module overlay, public-start change or existing-game retrofit is added.

Genuine first end-Mentat custody, all six effects, all five HMS copied advantages, current-control zero/sole/multiple choices, native CHOAM bank support and actual donor funding are connected. Eighteen native source-replay cases and all four projected legal policies pass. Five unstaged Advanced games (2..6 seats) complete with3,129 accepted actions, no rejections and82 JSON restorations. Earlier affected162/162 and full6,513/type/lint/build evidence was recorded before the user's rules-first assurance deferral; it is not a new per-feature gate.

Original local CLI room LAUDR2WN v14 continued its actual undealt setup without redealing. At390px the human Ixian chose Arrakeen from its real HMS copy choices, entered dial6/support3 with bank2/personal1, and sealed the native plan. The original opposed battle completed: personal1→0, three actual cyborg losses and native CHOAM income1. Refresh showed the real Collection opening, not a fictitious completed collection.

Fixture repairs use native CHOAM market done, phase-opening priority, legal Weirding pairing and actual Collection timing. A rare original private deal (initial seed12) used the real Guild-held Thumper for native Nexus alliance, without stealing/redealing it. Prior scoped reviews found no remaining defect. Save/privacy/custody assurance is deferred until all rules under the2 October priority; no new seat/database audit is scheduled. Ecaz rounding, Richese/cache/Stone and other overlays remain outside this profile. Complete factions/cards/modules and deployed acceptance remain open. Final focused guide/tooling25/25, check:quick, build and expanded390px Partial/DevelopmentVerified guide pass. No full-suite or assurance rerun was added after the user's priority amendment.

Authority and inventory

The original CHOAM & Richese module contains six separate Stronghold Cards, one of each. They never enter the Treachery deck, hands, discard, bidding pool or hand limit. The publisher rulebook requires Advanced play; neither CHOAM nor Richese needs to be a participating faction. The mobile card remains set aside when no placed mobile stronghold exists.

The user-authorized root Advanced2.3 PDF governs Advanced differences from 1 October onward; metadata2.3/editorial2.2 and physical-page provenance are in the source amendment. Printed faces and GF9 material supply details omitted there. The historical checkpoint's original sources are:

1. Publisher CHOAM & Richese rulebook, printed pp4, 6, 9–12: inventory, Advanced prerequisite, end-turn custody, CHOAM bank-funded support income and advisor timing. Readable publisher-authored mirror SHA-256 b628cef05b167953c2f192c8acfdefea92aee8f5497eea779edacf7be25e6299.
2. All six publisher-designed English physical faces, visually checked from the product-gallery photograph. Image, SHA-256 b489eaec185f0ecb3413a34276b19c93dfe852b37458dc436d3d976d015f7da4, 1024×768. This is a third-party photograph of the printed components; photographer, date and printing are unverified. It is research evidence, not a bundled asset.
3. Publisher Ecaz & Moritani rulebook, printed p8: Ecaz controls a stronghold occupied only by itself and its ally.
4. Classic base rules, pp10–11 and13; November2020 FAQ, pp2 and8: battle outcomes, support, dial meaning, Tooth and defenses. Later direct rulings take precedence over general older text. No tournament-rewritten Stronghold effects were adopted.

Printed identity	Canonical board ID	Complete effect in original language
Arrakeen	arrakeen	The bank contributes two spice toward actual force-support payment in battles there.
Carthag	carthag	Without a poison weapon in your plan, your played non-poison defense also functions as a Snooper.
Habbanya Sietch	habbanya_ridge_sietch	Win battle ties there regardless of storm order or Juice of Sapho.
Sietch Tabr	sietch_tabr	A win earns bank spice equal to the opposing dial rounded down.
Tuek’s Sietch	tueks_sietch	Each Worthless card played in the battle earns two bank spice; the face imposes no win condition.
Ixian Hidden Mobile Stronghold	hidden_mobile_stronghold	Copy an advantage of another controlled stronghold and declare it before Battle Plans.

No card has a printed once-per-turn limit. Benefits apply to their holder in that location, not globally. The copied benefit applies to the battle inside the mobile stronghold.

Custody and decisions

game/stronghold-cards.ts owns the canonical definitions, control query, costs and income arithmetic. Setup leaves every owner null. At final end-Mentat settlement, after eligible board changes and advisor normalization, each card transfers to its controller or is set aside. Cards remain with their recorded holder throughout the next turn, even after leaving, losing a battle or becoming advisors. claimedTurn prevents duplicate settlement. Normal turn completion and terminal victory both use the same ownership checkpoint.

Fighter presence across a territory counts; advisors do not. A concealed No-Field contributes one public presence without reading its private denomination. Ordinary allied armies do not pool control. Exactly Ecaz plus its reciprocal ally selects Ecaz, as the explicit faction exception. A mobile interior is distinct from its pointing territory.

Before mobile battle powers and plans, the card holder receives a public choice if multiple other strongholds are currently controlled. One eligible advantage is announced automatically; zero proceeds without a copied advantage. The persisted choice is bound to the battle event, owner, phase and exact eligible pool. An accepted copy remains fixed for this battle. Public views expose ownership, declared copy and each combatant’s effective advantage; sealed plans and foreign hands retain their existing privacy.

Composed-rule interpretations

These are recorded textual interpretations, not claims of dedicated publisher FAQ answers:

• Mobile control versus card custody: the face says controlled stronghold, whereas the general module separately permits retaining a departed location’s card. Therefore use current physical control, without requiring custody of that other card. A newly occupied location can qualify; an abandoned card alone cannot. No tournament end-of-Bidding timing or replacement copy pool is used.
• Arrakeen low costs: bank support is min(2, actual support). Unspent allowance never becomes personal cash, and free Fremen support cannot mint income. Normal and allied personal charges use the remaining cost. CHOAM’s published income includes the bank portion, excludes CHOAM’s own donor share, and follows the existing no-income-on-traitor rule.
• Carthag roles: defensive Weirding Way is projectile defense and can gain Snooper; defensive Chemistry already is poison defense. Worthless and empty slots cannot gain protection. A poison-role weapon, including Chemistry or Tooth in the weapon role, prevents this advantage. The added property does not replace physical identity: Shields still cause Lasgun explosions and Snooper protection does not stop Tooth. The defense actually used after any permitted late replacement supplies the property.
• Habbanya and Stone Burner: the general battle-tie advantage also decides a tied undialed-token comparison. It does not override a traitor victory or create a winner after mutual destruction. Sapho remains an unfinished card effect; this implementation does not certify its future ordering.
• Tabr: use the opposing declared dial, not leader strength, support cost or token count. A single traitor caller wins and qualifies. Explosion and double traitor have no winner and do not qualify.
• Tuek: ordinary defeat and single-traitor outcomes do not erase played Worthless cards. An explosion can still qualify. For double traitor, the base rule that neither player receives spice takes precedence, so neither receives this payout. CHOAM’s out-of-plan Worthless discard powers and BG Karama conversions are different uses and do not qualify.
• Ecaz ally battle: card custody belongs to Ecaz. No general transfer or automatic sharing of a held card’s benefit is inferred. The standalone E3 follow-up composes the original Advanced total/fixed/variable dial: only a selected Ecaz holder receives its advantage; selecting the ally does not lend Ecaz’s card.

Player experience and AI

The table has public ownership labels, a six-card gallery, an enlarged inspector, and a decision panel for mobile copies. Support controls include Arrakeen’s two bank spice and separate the actual bank contribution from player/ally payment. Successful automatic support, copy, ownership and income events use the existing house-colored notices without an acknowledgement.

The internal reference has searchable complete original effect prose, related phase/faction links, an example and five-part implementation checklist. It contains no external source links. Each illustration is original built-in image generation, saved locally under public/art/strongholds/*-v1.png; exact prompts, original paths, hashes and visual inspection notes are in STRONGHOLD_ART_GENERATION.json. Normal and enlarged faces retain crisp interface text independent of the artwork.

All four policies consume their entitled public copy/effect and private hand/support data. Arrakeen adds affordable supported half-step dial candidates so a large army does not overlook free support when its ideal dial is unaffordable. Carthag and Habbanya enter combat evaluation. Mobile choice heuristics consider funding, defense, defender priority and Worthless holdings. This is policy support, not proof of calibrated relative strength.

Verification

• 12 pure inventory/control/cost/income cases.
• 9 actual engine lifecycle and six-effect cases, including CHOAM donations, defender Stone tie, losing Worthless cards and stale mobile answers.
• 4 all-profile AI cases, including 4/6/10/20-force Arrakeen armies and hidden-state noninterference.
• 3 additional combat-role and mutual-effect matrix cases.
• 4 production-room SQLite tests: fresh module instances, real private-seat recovery, sealed-plan views, duplicate/crossed copy CAS, exactly-once subsidies/income, and eight corrupted saved-choice bindings. The room scheduler does not answer a pending human choice; direct engine normalization and actions reject invalid bindings without writing.
• Independent actual-action matrix: 17 Arrakeen/CHOAM support-share combinations, all paid once with the expected balances.
• 40 complete sampled base-faction Advanced games with the optional module: all player counts 2–6 × all four homogeneous profiles × two roster/seed replicates, nine distinct rosters. 19,665 accepted actions, zero rejected candidates/stalls/invariant failures, 224 setup actions and 912 continued JSON round trips. Observed 330 custody, 21 support and 7 income notices. Source hashes were stable during the 105.603-second run.

The game study uses the unchanged strict base audit initializer, then adds only empty Stronghold module configuration immediately before the first setup action. It does not change cards, resources, starting forces or the public start gate. Runner /tmp/dune-stronghold-fullgames.ts, full results /tmp/dune-stronghold-fullgames.json (SHA-256 d6f924aca8ed714df0a92008b4d6eda5b54d1e0192f4eda1a937dd98adbcf3e8), summary /tmp/dune-stronghold-fullgames-summary.json. This base-roster study does not exercise an Ixian interior or CHOAM full game; focused tests cover their bounded interactions.

The registered suites pass 1,528 rules/client/component tests and 133 persistence/API tests. Logs: /tmp/dune-stronghold-full.log and /tmp/dune-stronghold-multiplayer.log. Desktop normal and enlarged faces, plus all six enlarged phone-width faces at 390×844, were visually inspected with complete readable text and working close controls. Final live-control and build verification are recorded in the implementation status checkpoint.