Compose native Discovery skills and repair hidden trainer participation

This commit is contained in:
Quintine committed 2026-10-06 00:51:21 +11:00
1 parent d51142bf9a
commit 555afe4070
20 files changed
+2220 -27

No files matched your search

+32 -1
View File
@@ -1,6 +1,6 @@
# Current development status
Updated 5 October 2026. **The full twelve-faction goal is unfinished. Basic
Updated 6 October 2026. **The full twelve-faction goal is unfinished. Basic
is playable and the six classic factions have a user-authorized Advanced preview.
Expansion starts and full rules/publication acceptance remain gated.**
@@ -17,6 +17,37 @@ open mode gates.
## Current checkpoint and work
**Native Leader Skills / Discovery — bounded Development Verified:**
fresh supported native Basic/Advanced family predicates retain original
decks, starting hands/offers/all14, Discovery7/8, optional Tech3+ and Advanced
Strongholds2+. Typed entry/range/rescue/Cyborg/Face Dance, native stash/
invoice/green marker and trained Moritani death/assassination/own revival use
original handlers. Native Ecaz mandatory owner-labelled coalition remains.
Other overlays/previews, pending rulings, public starts and conversion stay gated.
Affected107 pass. Actual programmes observe rescue3/three Tanks/one Cyborg
substitution, six-counter invoice5→receiver3/bank2, and five real native
leader deaths followed by turn4 Revival/printed5 debit/optional replacement.
Three genuine hidden-Smuggler stalls had legal human plans; the policy now
uses authoritative projected capability, not full-state rules on player views.
Failing-before/passing-after trainer regression covers all four policies.
All31 genuine six-seat samples complete after three unchanged capture resumes:
30,874 accepted/30,874 attempts/815 JSON across f2ec1d69 and corrected14d66863
trees. Resumes add989/387/783 no-rejection actions; they do not reconstruct the
prior random stream. Earlier duplicate-name runs remain incomplete/un-counted;
new names include full roster/expansions while old profiles/seeds stay unchanged.
Human 6NNTCFM2v12 moves one ordinary/one Cyborg into Shrine free, preserving
natural Bureaucrat and HMS6. 5NLWDL68v14 keeps native Killer Medic: actual
Stash5→6→5 through new Karama disposal without payment. MYPGMSW6 human
Master Bewt revealv10 kills its naturally trained Guild Bureaucrat, returns
that skill and pays10→13; winner cleanupv11 and ten original actions reach
v12/Mentat with retired cost and one private replacement. All390px refresh,
original starting hands/offers preserved, explicit staging and tabs released.
[Canonical native evidence](DISCOVERY_PROTOTYPE.md#native-leader-skills-composition--6-october-2026).
No full-faction, natural-every-effect, assurance or deployed acceptance claim.
**Discovery / original Strongholds — bounded Development Verified:**
fresh original Discovery and separate classic Skills/Discovery preserve
Advanced Strongholds2–6, optional Tech3–6, original setup/held claims,
+78
View File
@@ -35,6 +35,84 @@ node --import tsx tools/start-prototype.ts --profile discovery --db PATH --room
It preserves all seats and other rooms, requires the current version and refuses
to redeal a started game. The default `ix` profile remains available.
## Native Leader Skills composition — 6 October 2026
Fresh explicit `leader-skills --discoveries` now preserves original native
family decks, all14 training, Discovery7/8 and each existing native Basic/
Advanced roster predicate. Original Tech needs3+; Strongholds requires Advanced.
Standalone Ecaz OR Moritani retains its Duke/assassination/Harkonnen exclusions.
No Nexus/Homeworlds/independent variants, Banker/Mentat or unrelated preview,
public start or played-save conversion is admitted.
The original native module predicates validate one Discovery envelope, not
copies of game/configuration objects. Ordinary skill quotes, Planetologist,
Suk/Rihani/Sandmaster, native invoices and Bureaucrat consume those predicates.
Original native starting hands and already offered training are authoritative;
some original E2/E3 constructors have dealt their one starting card before
training. Human programmes preserve that original unassigned setup rather
than redealing or demanding empty hands. No assigned/played setup is accepted.
Actual typed entry, paid nested routes, green components, rescue before real
Cyborg Tank substitution, cleanup/Tech before matching Face Dance, CHOAM5→6→5
Stash, native invoice5→receiver3/bank2, physical marker0/3/5 caps and mandatory
green cleanup compose existing handlers. Moritani nested assassination kills
the actual unused trained disc, returns its exact skill once, pays printed3
and receives one private actual Mentat replacement. Five original battles
kill all five native Moritani discs; original next Revival pays printed5
and opens optional replacement. No death, clock, wallet or earned receipt is
assigned to reach it.
### Native skill runtime evidence
The affected native/Smuggler union passes **107 cases**. Original rule programmes
observe three rescue survivors/three Tanks/one genuine Cyborg loss eligible
for substitution, the six-counter5-spice invoice with receiver3/bank2, and
five actual native disc deaths followed by turn4 Revival, printed5 debit,
four remaining dead discs and the private replacement decision.
Three genuine games stalled with a hidden Smuggler trainer after the opposing
plan. Actual human zero-support plans were legal, but bots incorrectly ran
full-state module predicates on partial player views. The authoritative battle
projection now supplies `smugglerCollectionSupported`; all four policies use
that quote while keeping the existing modified-strength/multiple-pile guards.
A real trainer-plan regression fails before repair and passes afterward.
The original captures resume at seeds20265473/20265497/20265501 and complete
989/387/783 more accepted actions without rejection and26/10/21 JSON.
All31 original six-seat samples now complete: nineteen Basic/Advanced native
Skills/Discovery/Tech and twelve Advanced four-module games. Combined original
and repaired continuations total30,874 accepted/30,874 attempts/815 JSON.
Initial reports bind unchanged
`f2ec1d696e1a5572abcb43a9e529b07cdf8e56dc7694f4c3c9f34cad3acd3958`;
the three corrected resumes bind unchanged
`14d6686355ebff299369e8c62f5b498e2194574b7a4eb69fc1ce5d9ff42668ea`.
This spans two trees; resume does not reconstruct the earlier stream.
Early duplicate scenario-name runs remain incomplete and are not counted:
native names now include the full original roster/selected expansion list,
without changing earlier profiles/defaults or new4400–4700 ordinal bands.
Actual **6NNTCFM2** original CLIv10 keeps natural Ixian Bureaucrat training.
Conserved original Shrine/counter positions and original first-Mentat claim
reachv11 entry. Human one-ordinary/one-Cyborg entry **v12** leaves Shrine2/
one Cyborg, one parent counter, unchanged spice/reserves and actual HMS6.
Actual **5NLWDL68** original CLIv10 keeps starting hands and natural CHOAM
Killer Medic. Explicit conserved native five-card capacity precedes Collection;
human inspectv12/revealv13 draws six, then human actual new Karama discard
**v14** restores five without price or training loss.
Actual **MYPGMSW6** original CLIv8 keeps native training and produced Cistern
after-loss opportunityv9. Human Master Bewt reveal **v10** spends the actual
Traitor, kills the Guild's naturally trained Bureaucrat, returns that skill
and pays10→13. Human winner cleanupv11 retains original Tech; ten original
actions finish the separate conflict/Collection and actual **v12/Mentat**
retires the cost and produces one private replacement. Native assassination
profile admission was repaired after its independent Discovery-first branch
rejected the real original setup. All three refresh at390px; tabs released.
Conserved positions, original token/card lotteries and original offers remain
explicit. Controlled deterministic programmes are not natural card histories;
passing samples do not prove natural use of every effect, complete factions,
full combination assurance, privacy/recovery certification or deployment.
## Classic Nexus composition — 5 October 2026
Fresh local `nexus --discoveries` composes classic/base33/all12 Nexus with the
+37
View File
@@ -1,5 +1,42 @@
# Dune implementation status
## 6 October 2026 — Original native Leader Skills and Discovery
One exact Discovery envelope feeds existing native family/roster predicates
and original all14 setup/starting hands, optional Tech3+/Advanced Strongholds2+.
No game/config copies in rule quotes; no public gate or save conversion.
Typed entry/range/rescue before Cyborg substitution, original cleanup/Tech
before Face Dance, CHOAM stash/invoice/Bureaucrat/green marker, and trained
Moritani nested death/assassination/private replacement/own revival compose
existing sources. Actual native E2/E3 printed starting hands are preserved.
Affected107 pass. Actual programmes observe rescue3/three Tanks/one real
Cyborg loss, six-counter5 fee/receiver3/bank2, and five original battle deaths
then turn4 native printed5 revival/optional replacement. Original chronology,
support-unit and separate-conflict fixture corrections remove copied math,
nonce/default assumptions rather than pinning implementation artifacts.
Three genuine hidden-Smuggler captures had legal human zero-support plans
but no bot candidate. Engine projects canonical support; bots stop recomputing
full-state module predicates on partial views. The real trainer-plan regression
fails before/passes after and keeps modified-strength/multi-pile guards.
Original seeds20265473/20265497/20265501 resume unchanged, adding989/387/783
accepted/no rejected actions and26/10/21 JSON on14d66863 tree.
All31 six-seat samples finish30,874/30,874/815 JSON including those resumes;
initial f2ec1d69 tree evidence is separate, not one frozen-tree/every-effect
claim. Earlier filename collisions are retained but un-counted; full original
rosters/expansions now distinguish new samples, without changing old names/seeds.
Actual 6NNTCFM2 original CLIv10, natural Bureaucrat/free typed Shrinev12;
5NLWDL68 original CLIv10, natural Killer Medic/Stash inspect12/reveal13/
new-card disposal14 restores5 with no payment; MYPGMSW6 original CLIv8,
real nested after-loss target9/human Master Bewt10 returns Guild Bureaucrat,
printed10→13, winner cleanup11 and original Mentat12 retires one actual cost/
draws one private replacement. Original hands/offers remain authoritative,
conserved positions explicit,390px refresh/tabs closed.
[Canonical native contract](DISCOVERY_PROTOTYPE.md#native-leader-skills-composition--6-october-2026).
Full modes, assurance, strategic AI and protected live deployment remain open.
## 5 October 2026 — Original Discovery Strongholds and mandatory Ecaz repair
Fresh original Discovery classic/E1/E2/standalone E3 and separate classic
+21
View File
@@ -5,6 +5,27 @@ Normal mode starts and publication remain gated. The [source contract](LEADER_SK
contains all fourteen physical cards; the [capture interpretation](LEADER_SKILLS_CAPTURE.md)
records the fresh publisher/designer search requested by the user.
## Original native Discovery composition — 6 October 2026
Fresh native `leader-skills --discoveries` composes each already-supported
Basic/Advanced native family predicate, original decks/hands/offers/all14 and
Discovery7/8, optional original Tech3+ and Advanced Strongholds2+.
E1 typed entry/range/rescue/Cyborg/Face Dance, E2 stash/invoice/green marker
and E3 trained death/assassination/own revival use original consumers.
Actual offered training and printed starting hands are not replaced.
The same supported live Ecaz profile keeps mandatory owner-labelled coalition
battles; shared Duke assignments and Basic odd-force rounding remain guarded.
Nexus/HW/variants/Banker/Mentat previews/public starts/save conversion stay
separate. This does not certify all fourteen effects or complete native factions.
[The Discovery contract](DISCOVERY_PROTOTYPE.md#native-leader-skills-composition--6-october-2026)
records affected107, actual producer programmes and three original human
390px paths. Thirty-one genuine six-seat samples complete after three actual
hidden-Smuggler policy captures are repaired/resumed:30,874 accepted/no
rejection/815 JSON across two frozen trees, not one unstaged every-effect proof.
Original Native starting-hand order, source-local held effects and physical
reward/skill-return ordering remain; comprehensive assurance stays deferred.
**2 October normal Banker checkpoint:** the separate fresh local `banker-income`
profile connects actual final-bank payment legs, one physical Banker phase
grant and automatic current Mentat collection. The full deck14
+1
View File
@@ -238,6 +238,7 @@ dedicated retrieved worked FAQs. [Contract and remaining limits](HOMEWORLD_OCCUP
| Paired E1/E2 Nexus / Leader Skills | Fresh both Ixians+Tleilaxu OR CHOAM+Richese plus classics, one original family deck/all14 Skills/all12 Nexus, Basic/Advanced2..6, explicit original skill entry. Actual native setup/offers/closing draws remain. Full unsupported Suboids compose typed Suk rescue before equal actual Tank substitution; Face Dance follows original winner outcomes, kills/returns the winning trainer skill once without another bounty, and partial Cunning retains ordinary Mentat rights. Richese pair uses its original signed two-token frame/one invoice/native response/Guild continuation, not single-marker skill proof; CHOAM printed fuel/Planetologist physical route remain separate. No other overlays, companion/mixed-plan/new ruling/public starts or retrofit. | [Scope, sources and bounded evidence](NEXUS_CARD_RULES.md#paired-e1e2-nexus-and-leader-skills-composition--4-october-2026), [Skill runtime](LEADER_SKILLS_RUNTIME.md#paired-e1e2-nexus-composition--4-october-2026) |
| Classic Nexus / Discovery / original components | Fresh classic/base33/all12/Discovery7/8 Basic/Advanced2–6, original Tech3+/Advanced Strongholds2+, explicit `nexus --discoveries`. Original worm/vote/typed ride, next-turn entry, end-Mentat held custody and end-Spice closing draw after a Nexus and settled alliance remain. Extra/ordinary revival ledgers and borrowed Guild nested payment preserve native identity and physical outcomes. Skills/native/HW/variants/previews and unresolved mixed-module Fremen Betrayal remain gated; no new universal interruption priority. | [Composition contract and bounded evidence](DISCOVERY_PROTOTYPE.md#classic-nexus-composition--5-october-2026), [Nexus source contract](NEXUS_CARD_RULES.md#classic-discovery--original-tech-and-strongholds--5-october-2026) |
| Discovery / original Advanced Strongholds | Fresh original Discovery classic/selected E1/E2/standalone E3 and separate classic all14 Skills entry retain Strongholds6 Advanced2–6, optional Tech3–6 and original end-Mentat claims. Named held benefits or actual HMS copy do not inherit into nested locations; Jacurutu creates no seventh card. Live supported Ecaz Discovery feeds mandatory original coalition lead/armies/payment/aftermath rather than skipping legal battle. Native typed/marker, trained rescue and winner Tech retain their sources. Basic Strongholds, native Discovery Skills, HW/variants/previews/pending rules/public starts/save conversion remain separate. | [Bounded contract and original programmes](DISCOVERY_PROTOTYPE.md#original-advanced-stronghold-composition--5-october-2026), [mandatory Ecaz repair](ECAZ_OCCUPY_RULES.md#original-live-discovery-coalition--5-october-2026) |
| Native Leader Skills / Discovery | Original supported native Basic/Advanced family/roster predicates retain selected native decks/hands/all14 and Discovery7/8, optional Tech3+/Advanced Strongholds2+. Typed rescue precedes actual Cyborg substitution; original cleanup/Tech precedes Face Dance. Native stash/invoice/green marker and trained assassination/own revival keep physical sources. Actual starting hands/offers are authoritative. Hidden-Smuggler legal participation uses the engine's projected support quote, not private-state profile reconstruction. Native Ecaz owner-labelled coalition remains; no shared-Duke training, new ruling, other overlays/public starts or conversion. | [Bounded native contract and original runtime/human evidence](DISCOVERY_PROTOTYPE.md#native-leader-skills-composition--6-october-2026), [Smuggler capture repair](SMUGGLER_BATTLE.md#native-discovery-legal-participation--6-october-2026) |
| Discoveries | Preserve seven sourced Spice Cards/eight tokens, faction peeks and Collection choices. Fresh classic/native E1/E2/standalone E3 may retain original Tech3..6; separate fresh classic all14 Skills/Discovery/Tech uses original training, entry, movement, stash, rescue/rewards and Bureaucrat invoices. Native setup, revealed-only No-Fields and Moritani source histories remain. Advanced Orgiz uses the supplied p24 territory-Collection trigger; Basic stays provisional per deposit. Only Jacurutu adds a stronghold. Contested/shared/mixed cases, other combinations and public starts remain unfinished. | [Component authority](DISCOVERY_COMPONENTS.md), [native composition](DISCOVERY_PROTOTYPE.md#native-setup-and-original-consumers--5-october-2026), [classic Skills/Advanced source cutover](DISCOVERY_PROTOTYPE.md#classic-leader-skills-and-advanced-orgiz--5-october-2026) |
| Guild Secret Ally | One explicit Guild-priced ordinary shipment; borrowed prices/routes preserve native identity. Free Fremen reinforcement remains a separate choice. | [Route contract](NEXUS_GUILD_SECRET_ALLY_RULES.md), [verified runtime](NEXUS_GUILD_SECRET_ALLY_RUNTIME.md) |
| Homeworld custody and mobility | Physical home/visitor groups, population thresholds, tariffs and transport permissions retain distinct sources; Homeworld immunity is not universal immunity. | [Homeworld rules](HOMEWORLD_RULES.md), [benefits](HOMEWORLD_BENEFITS_RULES.md), [mobility](HOMEWORLD_MOBILITY_RULES.md), [invasion](HOMEWORLD_INVASION_RULES.md), [Guild transport](GUILD_HOMEWORLD_TRANSPORT_RULES.md) |
+13
View File
@@ -4,6 +4,19 @@
the base-faction Leader Skills preview connect the lower Smuggler effect in Basic
and Advanced games. Public module, mode and publication gates remain closed.
## Native Discovery legal participation — 6 October 2026
Three genuine native Skills/Discovery/Tech captures had legal hidden-trainer
plans but no bot candidate: the policy recomputed full-state module admission
against a player view that intentionally lacks private/internal configuration.
The engine now projects `smugglerCollectionSupported` beside the existing
saved battle format flag; bots use that authoritative rule result rather than
reconstructing the module profile. Modified-strength and multiple-positive-pile
guards still apply to each actual plan. No public mode or strategy tuning is
added. A real trainer-plan regression fails before and passes after repair;
all three original captures finish without rejection. Source trees/counts and
limits are in the [native Discovery evidence](DISCOVERY_PROTOTYPE.md#native-skill-runtime-evidence).
## Rule and timing contract
The [source follow-up](LEADER_SKILLS_RULES.md#20-september-smuggler-follow-up)