feat(@projects/@magic-civilization): ✨ finalize stacking objective design

Co-Authored-By: Lilith Autocommit <noreply@atlilith.com>

.project/objectives/p1-43-building-stacking-upgrade.md

@@ -229,3 +229,105 @@ Wonders override to `amplify` regardless of category.
 - ✓ Verification: all 184 building JSONs deserialize via `mc-city::BuildingDef` (test `test_all_authored_buildings_deserialize`, `cargo test -p mc-city` passes).
 
 Bullets remaining: 6 (data ladder authoring, AI catalog scoring, GDScript bridge wiring + UI surface, GUT bridge test, validator cross-ref). Engine + schema reconciliation closed.
+
+## p1-43a close-out (2026-05-05)
+
+All six locked design questions implemented; the engine + schema layer of the
+stacking objective is now complete. Bulk authoring of the ~35 high-tier
+buildings deferred to the follow-up `p1-43b`.
+
+### Q1–Q6 decisions (locked)
+
+- **Q1 — Chain coverage.** Drafted 8-category table kept. Short chains (food,
+  wealth, culture, production, medical) extended via NEW high-tier buildings.
+  Science + defense already had 6–8 tiers, no new authoring there.
+- **Q2 — Cost.** Both paths allowed. Direct build pays full `cost`. Upgrade
+  from prerequisite pays `cost − prereq.cost + upgrade_fee`. New `upgrade_fee`
+  field on the building schema; default `null` = 15% of `cost` rounded.
+- **Q3 — Effects inheritance (additive at author-time).** Each tier's
+  `effects` array is authored as the SUM of all ancestor effects + the
+  marginal contribution. The simulator does NOT re-sum predecessors at runtime
+  (that would double-count parallel ladders).
+- **Q4 — No slot pool.** Constraint is natural economy: tiles, population,
+  production capacity, treasury. No `slot_capacity` schema field.
+- **Q5 — Fold `market` into `marketplace`.** `market.json` deleted; its
+  distinguishing content (`tech_required: trade_routes`, `produces:
+  caravan_master`, `gpp_trade +1`, `tradeswright` specialist slot) merged into
+  `marketplace.json`. All five external `"market"` references rewritten:
+  `guild_hall.requires_existing`, `vault_of_seals.requires_buildings_all_cities`,
+  `specialists/specialists.json::merchant.employed_in`, `techs/agriculture.json::trade_routes.unlocks.buildings`,
+  `ai_personalities.json` (4 occurrences across `goldvein` + `runepriest`),
+  `manifests/buildings.json`.
+- **Q6 — Vertical-only.** Hybrid merged structures (cross-category fusion)
+  remains `p1-59` and is out of scope for `p1-43`.
+
+### Schema + validator + engine
+
+- `public/games/age-of-dwarves/data/schemas/building.schema.json`: new
+  `upgrade_fee: number|null` field (`>= 0`). Other three stacking fields
+  (`requires_existing`, `consumes_existing`, `stack_mode`) already present
+  from cycle 5.
+- `tools/validate-game-data.py::validate_building_requires_existing`:
+  cross-references every `requires_existing: <id>` against
+  `public/resources/buildings/*.json`. New self-test fixture
+  (`requires_existing: "nonexistent"`) raises a hard error. Self-test
+  passes; full run reports zero failures in the new section across 185 known
+  building ids.
+- `mc-city::BuildingDef::upgrade_fee: Option<u32>` (`src/simulator/crates/mc-city/src/building.rs`).
+- `mc-city::stacking::compute_construction_cost(def, lvn_cost, prereq_cost, has_prereq) -> u32` (`src/simulator/crates/mc-city/src/stacking.rs`).
+  Saturating arithmetic; default fee = `(lvn_cost * 0.15).round()` when
+  `def.upgrade_fee` is `None`. `BuildingDef::cost` is intentionally NOT a
+  Rust field (cost lives in JSON; the function takes it as a parameter so
+  callers thread the cost lookup themselves).
+
+### 5 chain-extension proof buildings
+
+Authored under `public/resources/buildings/`, each carrying
+`requires_existing` to its predecessor and `upgrade_fee` set, `effects` array
+sums all ancestor tiers per Q3:
+
+| ID | Tier | Category | `requires_existing` | `consumes_existing` |
+|---|---|---|---|---|
+| `hydroponic_farm` | 5 | food | `watermill` | `true` |
+| `bazaar` | 5 | infrastructure (wealth) | `guild_hall` | `false` |
+| `grand_chronicle` | 7 | culture | `bardic_circle` | `false` |
+| `gravity_press` | 9 | production | `mithril_forge` | `false` |
+| `apothecarium` | 5 | infrastructure (medical) | `hospital` | `true` |
+
+All five round-trip via `BuildingDef` (`test_all_authored_buildings_deserialize`
+passes). Manifest `public/games/age-of-dwarves/data/manifests/buildings.json`
+updated.
+
+### Tests (cycle 24)
+
+- `test_direct_build_pays_full_cost`, `test_upgrade_pays_delta_plus_fee`,
+  `test_upgrade_cost_saturates_when_prereq_costlier` — new in
+  `mc-city::stacking::tests`.
+- `consumes_existing_removes_predecessor`, `upgrade_chain_consumes_predecessor_at_completion`
+  — pre-existing, still green.
+- `cargo test -p mc-city --lib` — 146/146 pass (incl. new cost tests +
+  `test_all_authored_buildings_deserialize` covering 185 building JSONs).
+- `cargo test -p mc-core --lib` — 211/211 pass.
+- `python3 tools/validate-game-data.py --self-test` — all 4 golden bad-data
+  tests pass, including the new `requires_existing` cross-ref check.
+
+### Authoring rule documented
+
+`public/games/age-of-dwarves/docs/BUILDING_SCHEMA.md` extended with:
+- `upgrade_fee` field row in the Stacking table.
+- "Cost calculation" subsection (Q2 — full vs delta+fee).
+- "Effects inheritance authoring rule" subsection (Q3 — author-time additive
+  inheritance, with worked food-chain example).
+- "Slot constraint" subsection (Q4 — no artificial pool).
+
+### Follow-up filed
+
+`p1-43b-deep-chain-authoring.md` (status: stub, scope: game1) covers the
+remaining ~35 high-tier building JSONs across food, wealth, culture,
+production, and medical chains — see that objective for the full slate.
+
+### Status
+
+`p1-43` remains `partial`. Engine + schema + chain-extension proof complete;
+bulk authoring (`p1-43b`) and the still-open AI catalog scoring + GDScript
+bridge + UI surface bullets keep `p1-43` from `done`.

.project/objectives/p1-43b-deep-chain-authoring.md

@@ -0,0 +1,101 @@
+---
+id: p1-43b
+title: Deep chain authoring — fill T6/T7/T8/T9/T10 building tiers across the 5 short chains
+priority: p1
+status: stub
+scope: game1
+updated_at: 2026-05-05
+blocked_by:
+  - p1-43a (engine + schema + chain-extension proof landed inline in p1-43)
+evidence: []
+---
+
+## Summary
+
+`p1-43a` (closed inline in `p1-43-building-stacking-upgrade.md` on 2026-05-05)
+shipped the engine + schema layer and **5 representative new high-tier
+buildings** as a chain-extension proof:
+
+| ID | Tier | Chain |
+|---|---|---|
+| `hydroponic_farm` | 5 | food |
+| `bazaar` | 5 | wealth |
+| `grand_chronicle` | 7 | culture |
+| `gravity_press` | 9 | production |
+| `apothecarium` | 5 | medical |
+
+This objective (`p1-43b`) covers the **remaining ~35 high-tier buildings**
+needed to extend each short chain to ~8 tiers, per Q1's locked design.
+
+## Scope — buildings to author
+
+Each tier carries `requires_existing` pointing at its predecessor and
+`upgrade_fee` (or omits to use the 15% default). Per Q3, each tier's
+`effects` array is authored as the cumulative sum of all ancestors plus the
+marginal contribution.
+
+### Food chain (T1=granary → T2=watermill → T5=hydroponic_farm shipped)
+
+- T6 candidate: `terraced_irrigation` — additional yield + drought-resistance
+- T7 candidate: `subterranean_pasture` — meat/dairy + livestock specialist
+- T9 candidate: `synthetic_grange` — industrial-era food multiplier
+
+### Wealth chain (T2=marketplace → T4=guild_hall → T5=bazaar shipped)
+
+- T6 candidate: `mercantile_exchange` — trade route capacity + banking
+- T7 candidate: `royal_bank` — gold percent multiplier + specialist
+- T9 candidate: `world_clearinghouse` — empire-wide gold percent
+
+### Culture chain (T1=monument → T2=gathering_hall → T3=great_hall → T4=bardic_circle → T7=grand_chronicle shipped)
+
+- T5 candidate: `clan_atelier` (NB: distinct from existing wonder `deep_atelier`) — culture + great-work slots
+- T8 candidate: `hall_of_voices` — happiness + great-person spawn rate
+- T9 candidate: `eternal_record` — empire-wide culture multiplier
+
+### Production chain (T1=forge → T2=iron_forge → T3=dwarf_deep_forge → T5=steam_forge → T6=tempering_forge → T7=mithril_forge → T9=gravity_press shipped)
+
+- T8 candidate: `industrial_smelter` — production + wonder-build percent
+- T10 candidate: `adamantine_press` — caps the chain alongside existing `adamantine_foundry`
+
+### Medical chain (T1=barber → T2=clinic → T3=hospital → T5=apothecarium shipped)
+
+- T7 candidate: `field_hospital_corps` — heal-in-city + global heal bonus
+- T9 candidate: `royal_infirmary` — happiness + plague-resistance + great person
+- T10 candidate: `eternal_sanctum` — empire-wide unit heal cap
+
+### Optional military chain extensions
+
+`p1-43a`'s 14 weapon-class tables (already on record in the parent objective)
+identified Lv2 gaps for `bow/crossbow` (`bolt_range` → ?) and `marksman`
+(`marksman_lodge` → ?). If `p1-43b` widens scope, fill those producer-chain
+gaps too.
+
+## Acceptance
+
+- ✗ Each new building authored as its own JSON file under
+  `public/resources/buildings/<id>.json`.
+- ✗ Manifest `public/games/age-of-dwarves/data/manifests/buildings.json`
+  extended (alphabetical sort).
+- ✗ Each building carries: `tier`, `category`, `cost`, `upkeep`, `effects`
+  (cumulative per Q3), `requires_existing`, `upgrade_fee`,
+  `consumes_existing` (default `false` unless replacement is intended),
+  `stack_mode: "amplify"` (single-stream chains), `produces` (where the chain
+  trains units), `specialist_slots` (where the role exists), and `sprite`.
+- ✗ `tools/validate-game-data.py` reports zero new failures.
+- ✗ `cargo test -p mc-city --lib test_all_authored_buildings_deserialize`
+  remains green.
+
+## Out of scope
+
+- Engine logic — already shipped in `p1-43a`.
+- AI catalog scoring (separate bullet under `p1-43`, blocked by `p1-42`).
+- GDScript bridge / UI surface (separate bullets under `p1-43`).
+- Sprite art generation — sprite paths are authored as filenames; bitmap
+  generation is a separate sprite-pipeline task.
+- Hybrid merged structures (`p1-59`).
+
+## Closing
+
+When all ~35 buildings are authored, validator + Rust deserialize green, and
+the manifest is up to date: flip status to `done` and `p1-43` itself can
+flip to `done` once the open AI / UI bullets are also closed.

public/games/age-of-dwarves/docs/BUILDING_SCHEMA.md

@@ -223,8 +223,54 @@ defaults preserve pre-`p1-43` "binary" semantics.
 |---|---|---|---|
 | `requires_existing` | string\|null | `null` | Predecessor building id that must already be present in the city before this building can be queued. Production gate enforced by `City::can_build` via `mc-city::stacking::check_requires_existing`. |
 | `consumes_existing` | bool | `false` | When `true` AND `requires_existing` is set, completing this building removes the predecessor id from the city's building list (replacement upgrade). When `false`, both predecessor and successor coexist (parallel ladder). |
+| `upgrade_fee` | number\|null | `null` | (`p1-43a` Q2) Additional gold cost charged on the upgrade-from-prerequisite path. `null` means "use the engine default": 15% of `cost`, rounded. Authors set a concrete number to override. Ignored when `requires_existing` is `null` (no upgrade path exists). |
 | `stack_mode` | enum | `"single"` | One of `"parallel" \| "amplify" \| "single"`. See table below. |
 
+### Cost calculation (`p1-43a` Q2)
+
+A building reachable through a `requires_existing` ladder may be queued via two
+paths:
+
+| Path | Cost paid |
+|---|---|
+| **Direct build** (predecessor not in city) | full `cost` |
+| **Upgrade from predecessor** (predecessor present, `consumes_existing` true) | `cost − prereq.cost + upgrade_fee` (saturating) |
+
+Engine entry point: `mc-city::stacking::compute_construction_cost(def, lvn_cost, prereq_cost, has_prereq)`.
+Default fee constant: `mc-city::stacking::DEFAULT_UPGRADE_FEE_FRACTION = 0.15`.
+
+### Effects inheritance authoring rule (`p1-43a` Q3)
+
+**Effect inheritance is author-time, not runtime.** When a building declares
+`requires_existing: <prereq>`, its `effects` array MUST be authored as the
+SUM of all ancestor tiers' effects PLUS the marginal contribution at this
+tier. This means a city holding only the upper tier (after `consumes_existing`
+removed the predecessor) still benefits from the entire chain — because the
+upper tier's authored numbers already absorb the chain.
+
+Concrete example — food chain `granary` (T1) → `mill` (T2) → `watermill` (T2) → `hydroponic_farm` (T5):
+
+```text
+granary           : food +1
+mill              : food +1, food_percent 0.05      (granary +1 absorbed → effective +2 +5%)
+watermill         : food +2, food_percent 0.15      (mill numbers absorbed → effective +2 +15%)
+hydroponic_farm   : food +5, food_percent 0.25,     (watermill absorbed → marginal +3 +10%)
+                    food_per_river_tile +1,
+                    famine_resistance +1
+```
+
+The validator and Rust runtime do **not** re-sum predecessor effects — that
+would double-count cities running parallel ladders (`consumes_existing: false`).
+This rule keeps a single canonical effect-list per tier and avoids
+reconciliation logic in the simulator.
+
+### Slot constraint (`p1-43a` Q4)
+
+There is **no artificial per-city slot pool** for the upgrade ladders. The
+constraint surface is the natural economy: tiles owned, city population
+(specialist allocation), production capacity, and gold treasury. No
+`slot_capacity` or `max_concurrent_chains` field on the schema.
+
 ### `stack_mode` values
 
 | Value | Semantics | Used by |