diff --git a/docs/compiler/pbs/agendas/18.5. Ignored Call Results in Executable Lowering Agenda.md b/docs/compiler/pbs/agendas/18.5. Ignored Call Results in Executable Lowering Agenda.md
new file mode 100644
index 00000000..fae03003
--- /dev/null
+++ b/docs/compiler/pbs/agendas/18.5. Ignored Call Results in Executable Lowering Agenda.md	
@@ -0,0 +1,127 @@
+# Ignored Call Results in Executable Lowering Agenda
+
+## Status
+
+Open
+
+## Domain Owner
+
+`docs/compiler/pbs`
+
+Este tema pertence ao domínio PBS do compiler porque afeta:
+
+- semântica prática de `expression statement`;
+- lowering executável para `IRBackend`/`IRVM`;
+- ergonomia de uso de SDKs e hosts que retornam status;
+- compatibilidade entre o que o código-fonte parece permitir e o que o pipeline realmente aceita.
+
+## Problema
+
+Hoje uma chamada com retorno usada como statement isolado pode deixar valor sobrando na stack no lowering executável.
+
+Exemplo real:
+
+```pbs
+Gfx.set_sprite(...);
+```
+
+No estado atual, isso compila no frontend, mas pode falhar no pipeline backend com erro de validação de stack no `RET` da função, porque o resultado da call não foi consumido explicitamente.
+
+Isso cria uma inconsistência operacional:
+
+1. o código parece válido para o autor;
+2. o frontend aceita a forma;
+3. o backend exige, na prática, um consumo manual do retorno, por exemplo:
+
+```pbs
+let sprite_status = Gfx.set_sprite(...);
+```
+
+## Contexto
+
+O problema apareceu ao migrar o consumo de sprite em PBS para o contrato novo de `Gfx.set_sprite`, que retorna um status inteiro.
+
+Os pontos observados no código atual são:
+
+- `ExpressionStatement` faz lowering apenas da expressão;
+- o lowering de `CallExpr` emite `CALL_HOST`, `CALL_FUNC` ou `CALL_INTRINSIC` com `retSlots`;
+- não há descarte automático do valor quando a expressão é usada apenas como statement;
+- o validador de `IRVM` exige que funções `void` retornem com stack height `0`.
+
+Na prática, isso transforma "ignorar retorno" em comportamento parcialmente suportado pela linguagem, mas não suportado pelo pipeline executável.
+
+## Opções
+
+### Opção A
+
+Manter o comportamento atual e exigir consumo explícito de todo retorno em PBS.
+
+### Opção B
+
+Permitir que `expression statements` descartem automaticamente o resultado quando a expressão produzir valor.
+
+### Opção C
+
+Permitir descarte automático apenas para um subconjunto de calls, como hosts/SDKs anotados como status descartável.
+
+## Tradeoffs
+
+### Opção A
+
+- Prós:
+  - regra simples no backend;
+  - evita descarte implícito sem intenção.
+- Contras:
+  - ergonomia ruim para APIs baseadas em status;
+  - surpresa para autores, porque `foo();` parece natural mas falha mais tarde;
+  - expõe detalhe de stack do backend ao código-fonte.
+
+### Opção B
+
+- Prós:
+  - comportamento esperado para statement de expressão;
+  - reduz ruído em código de jogo e SDK;
+  - alinha parsing, semântica prática e lowering executável.
+- Contras:
+  - introduz descarte implícito;
+  - exige regra clara para saber quando emitir `POP`.
+
+### Opção C
+
+- Prós:
+  - mais controle semântico;
+  - evita descarte implícito amplo.
+- Contras:
+  - adiciona complexidade de contrato e metadata;
+  - mistura política de produto/API dentro do lowering;
+  - não resolve a expectativa geral sobre `expression statement`.
+
+## Recomendação
+
+Seguir com a **Opção B**.
+
+Direção recomendada:
+
+1. `expression statement` deve ser permitido mesmo quando a expressão produzir valor;
+2. o lowering executável deve emitir descarte explícito do resultado não usado;
+3. a regra deve ser geral para expressões com resultado materializado em stack, não especial para `Gfx.set_sprite`;
+4. testes de regressão devem cobrir host calls, callable calls e intrinsics com retorno ignorado.
+
+Essa direção preserva a ergonomia esperada da linguagem e remove um vazamento indevido do modelo de stack do backend para o código PBS.
+
+## Perguntas em Aberto
+
+1. O descarte automático deve valer para qualquer `expression statement` com valor ou apenas para formas lowerables em v1?
+2. O frontend semântico deve emitir algum warning opcional para retorno ignorado, ou isso fica fora do escopo atual?
+3. O descarte deve acontecer somente no lowering executável ou também virar regra explícita de spec para statements?
+4. Existe algum caso em que o descarte automático possa mascarar bug real de usuário que hoje seria detectado mais cedo?
+
+## Próximo Passo Sugerido
+
+Converter esta agenda em uma `decision` do domínio `compiler/pbs` fechando a política para `expression statement` com resultado ignorado.
+
+Depois disso, abrir um `pull-request/plan` curto para:
+
+1. ajustar o lowering de `ExpressionStatement`;
+2. adicionar fixtures e testes de regressão no frontend/backend pipeline;
+3. propagar a regra para specs relevantes de statements e lowering.
diff --git a/docs/compiler/pbs/agendas/README.md b/docs/compiler/pbs/agendas/README.md
index f2808ef7..8a9ad6cb 100644
--- a/docs/compiler/pbs/agendas/README.md
+++ b/docs/compiler/pbs/agendas/README.md
@@ -11,6 +11,7 @@ Closed agendas are moved to `docs/pbs/agendas/archive`.
 3. `18.2. Backend Workshop 2 - LowerToIRVM and IRVM Contract.md`
 4. `18.3. Backend Workshop 3 - Bytecode Marshaling and Runtime Conformance.md`
 5. `18.4. Asset References in Game Code - Names vs Compile-Time Lowering Agenda.md`
+6. `18.5. Ignored Call Results in Executable Lowering Agenda.md`
 
 ## Purpose
 
diff --git a/docs/packer/agendas/README.md b/docs/packer/agendas/README.md
index cfc82d71..73911669 100644
--- a/docs/packer/agendas/README.md
+++ b/docs/packer/agendas/README.md
@@ -7,6 +7,7 @@ This directory contains active packer discussion agendas.
 1. [`Tilemap and Metatile Runtime Binary Layout Agenda.md`](./Tilemap%20and%20Metatile%20Runtime%20Binary%20Layout%20Agenda.md)
 2. [`Pack Wizard Studio Decision Propagation Agenda.md`](./Pack%20Wizard%20Studio%20Decision%20Propagation%20Agenda.md)
 3. [`Tile Bank Packing Materialization Agenda.md`](./Tile%20Bank%20Packing%20Materialization%20Agenda.md)
+4. [`Variable Tile Bank Palette Serialization Agenda.md`](./Variable%20Tile%20Bank%20Palette%20Serialization%20Agenda.md)
 
 The first packer agenda wave was consolidated into normative specs under [`../specs/`](../specs/) and didactic material under [`../learn/`](../learn/).
 
diff --git a/docs/packer/agendas/Variable Tile Bank Palette Serialization Agenda.md b/docs/packer/agendas/Variable Tile Bank Palette Serialization Agenda.md
new file mode 100644
index 00000000..bb1a1c9a
--- /dev/null
+++ b/docs/packer/agendas/Variable Tile Bank Palette Serialization Agenda.md	
@@ -0,0 +1,142 @@
+# Variable Tile Bank Palette Serialization Agenda
+
+## Status
+
+Active
+
+## Domain Owner
+
+- `docs/packer`
+
+Cross-domain dependency:
+
+- `../runtime`
+
+## Purpose
+
+Decide whether `tile bank` payloads should keep serializing a fixed block of `64` palettes in v1, or evolve to serialize only the palettes actually authored by the asset.
+
+## Context
+
+The current `tile bank` contract is aligned between `packer` and `runtime`:
+
+- pixels are serialized as packed `u4`;
+- palettes are serialized as `RGB565 u16`;
+- `palette_count = 64`;
+- payload size is derived from `ceil(width * height / 2) + (64 * 16 * 2)`.
+
+This means a `tile bank` with `1` authored palette still serializes `64` palette slots. Unused slots are currently zero-filled, which is valid for the runtime, but wastes cartridge space.
+
+The packer now exposes:
+
+- `metadata.palette_count = 64`
+- `metadata.palette_authored = <actual authored count>`
+
+This makes the waste explicit, but does not remove it.
+
+## Problem
+
+The runtime-facing v1 contract favors fixed-size decode simplicity over cartridge efficiency.
+
+We need to decide whether:
+
+1. fixed `64` palette slots remain the normative contract for `tile bank`;
+2. `palette_count` becomes variable and payload size shrinks to the authored set;
+3. another staged strategy is better.
+
+## Options
+
+### Option A: Keep Fixed 64-Palette Serialization
+
+Keep the current contract unchanged:
+
+- `palette_count` remains `64`;
+- every `tile bank` serializes `64 * 16 * 2 = 2048` bytes of palettes;
+- authored palettes populate their indexed slots;
+- unused slots are zero-filled.
+
+### Option B: Move to Variable Palette Serialization
+
+Evolve the contract so `tile bank` serializes only authored palettes:
+
+- `palette_count` becomes the authored count;
+- payload size becomes `ceil(width * height / 2) + (palette_count * 16 * 2)`;
+- runtime validates and decodes variable palette blocks.
+
+### Option C: Keep V1 Fixed, Design Variable Serialization for V2
+
+Preserve the current runtime-facing v1 contract now, but explicitly open a v2 format/contract:
+
+- v1 stays fixed at `64`;
+- v2 introduces variable palette serialization;
+- packer and runtime keep clear compatibility boundaries.
+
+## Tradeoffs
+
+### Option A
+
+Pros:
+
+- no further runtime work;
+- simplest decode path;
+- current specs and tests stay valid.
+
+Cons:
+
+- every `tile bank` pays the full 2 KB palette cost;
+- cartridge size is inflated for small authored palette sets;
+- the metadata now exposes a mismatch between authored and serialized counts by design.
+
+### Option B
+
+Pros:
+
+- reduces cartridge size;
+- makes `palette_count` semantically tighter;
+- removes fixed empty palette padding from the payload.
+
+Cons:
+
+- requires coordinated changes in `packer`, `runtime`, specs, and tests;
+- changes the `size` / `decoded_size` formulas;
+- raises compatibility and migration questions for existing assets and cartridges.
+
+### Option C
+
+Pros:
+
+- keeps current work stable;
+- makes room for a cleaner future contract;
+- avoids mixing optimization work into the current tile-bank rollout.
+
+Cons:
+
+- known waste remains in v1;
+- adds future format/version management work.
+
+## Recommendation
+
+Recommendation for now: `Option C`.
+
+Reasoning:
+
+- the current fixed-size contract is already implemented and aligned;
+- the waste is real, but it is not a correctness bug;
+- changing it now would reopen a cross-domain runtime contract that was just stabilized;
+- if we want variable palette serialization, it should be introduced deliberately as a versioned follow-up, not as an incidental tweak.
+
+## Open Questions
+
+1. Should variable palette serialization be introduced as a new `tile bank` payload version, or as an incompatible change to the current one?
+2. If `palette_count` becomes variable, should runtime still materialize a 64-slot bank in memory, or truly shrink the resident representation too?
+3. Should palette slot identity remain sparse by authored `index`, or should serialization canonicalize into a dense runtime block?
+4. Which domain owns the compatibility story for existing cartridges: `packer`, `runtime`, or `shipper`?
+
+## Expected Follow-up
+
+If the team wants to eliminate the fixed 64-palette padding:
+
+1. create a corresponding runtime agenda/decision;
+2. define the versioning and compatibility story;
+3. update `packer` and `runtime` specs together;
+4. then plan code changes in both domains.
diff --git a/prometeu-compiler/frontends/prometeu-frontend-pbs/src/main/resources/stdlib/1/sdk/gfx/main.pbs b/prometeu-compiler/frontends/prometeu-frontend-pbs/src/main/resources/stdlib/1/sdk/gfx/main.pbs
index a3f6edd9..8fb80fea 100644
--- a/prometeu-compiler/frontends/prometeu-frontend-pbs/src/main/resources/stdlib/1/sdk/gfx/main.pbs
+++ b/prometeu-compiler/frontends/prometeu-frontend-pbs/src/main/resources/stdlib/1/sdk/gfx/main.pbs
@@ -28,7 +28,7 @@ declare host LowGfx {
   [Host(module = "gfx", name = "set_sprite", version = 1)]
   [Capability(name = "gfx")]
   fn set_sprite(
-    asset_name: str,
+    bank_id: int,
     index: int,
     x: int,
     y: int,
@@ -82,7 +82,7 @@ declare service Gfx
     }
 
     fn set_sprite(
-        asset_name: str,
+        bank_id: int,
         index: int,
         x: int,
         y: int,
@@ -94,7 +94,7 @@ declare service Gfx
         priority: int
     ) -> int
     {
-        return LowGfx.set_sprite(asset_name, index, x, y, tile_id, palette_id, active, flip_x, flip_y, priority);
+        return LowGfx.set_sprite(bank_id, index, x, y, tile_id, palette_id, active, flip_x, flip_y, priority);
     }
 
     fn draw_text(x: int, y: int, message: str, color: Color) -> void
diff --git a/prometeu-compiler/frontends/prometeu-frontend-pbs/src/test/java/p/studio/compiler/services/PBSFrontendPhaseServiceTest.java b/prometeu-compiler/frontends/prometeu-frontend-pbs/src/test/java/p/studio/compiler/services/PBSFrontendPhaseServiceTest.java
index cba2b53e..432df120 100644
--- a/prometeu-compiler/frontends/prometeu-frontend-pbs/src/test/java/p/studio/compiler/services/PBSFrontendPhaseServiceTest.java
+++ b/prometeu-compiler/frontends/prometeu-frontend-pbs/src/test/java/p/studio/compiler/services/PBSFrontendPhaseServiceTest.java
@@ -627,6 +627,57 @@ class PBSFrontendPhaseServiceTest {
         assertTrue(irBackend.getExecutableFunctions().stream().anyMatch(function -> "frame".equals(function.callableName())));
     }
 
+    @Test
+    void shouldLowerSdkGfxSetSpriteFacadeUsingBankIdContract() throws IOException {
+        final var projectRoot = tempDir.resolve("project-bootstrap-sdk-gfx-set-sprite");
+        final var sourceRoot = projectRoot.resolve("src");
+        final var modulePath = sourceRoot.resolve("app");
+        Files.createDirectories(modulePath);
+
+        final var sourceFile = modulePath.resolve("source.pbs");
+        final var modBarrel = modulePath.resolve("mod.barrel");
+        Files.writeString(sourceFile, """
+                import { Gfx } from @sdk:gfx;
+
+                fn frame() -> int
+                {
+                    return Gfx.set_sprite(2, 0, 12, 18, 7, 3, true, false, true, 1);
+                }
+                """);
+        Files.writeString(modBarrel, "pub fn frame() -> int;");
+
+        final var projectTable = new ProjectTable();
+        final var fileTable = new FileTable(1);
+        final var projectId = projectTable.register(ProjectDescriptor.builder()
+                .rootPath(projectRoot)
+                .name("app")
+                .version("1.0.0")
+                .sourceRoots(ReadOnlyList.wrap(List.of(sourceRoot)))
+                .build());
+
+        registerFile(projectId, projectRoot, sourceFile, fileTable);
+        registerFile(projectId, projectRoot, modBarrel, fileTable);
+
+        final var ctx = new FrontendPhaseContext(
+                projectTable,
+                fileTable,
+                new BuildStack(ReadOnlyList.wrap(List.of(projectId))),
+                1);
+        final var diagnostics = DiagnosticSink.empty();
+
+        final var irBackend = new PBSFrontendPhaseService().compile(
+                ctx,
+                diagnostics,
+                LogAggregator.empty(),
+                BuildingIssueSink.empty());
+
+        assertTrue(diagnostics.stream().noneMatch(d ->
+                d.getCode().equals(PbsSemanticsErrors.E_SEM_EXEC_LOWERING_UNRESOLVED_CALLEE.name())));
+        assertTrue(irBackend.getExecutableFunctions().stream().anyMatch(function -> "frame".equals(function.callableName())));
+        assertTrue(irBackend.getReservedMetadata().hostMethodBindings().stream()
+                .anyMatch(h -> h.ownerName().equals("LowGfx") && h.sourceMethodName().equals("set_sprite")));
+    }
+
     @Test
     void shouldReportInconsistentBarrelInsideSdkBootstrapFixture() throws IOException {
         final var projectRoot = tempDir.resolve("project-stdlib-barrel-inconsistent");
diff --git a/prometeu-compiler/prometeu-build-pipeline/build.gradle.kts b/prometeu-compiler/prometeu-build-pipeline/build.gradle.kts
index 393bd0e4..c590da1f 100644
--- a/prometeu-compiler/prometeu-build-pipeline/build.gradle.kts
+++ b/prometeu-compiler/prometeu-build-pipeline/build.gradle.kts
@@ -8,4 +8,12 @@ dependencies {
     implementation(project(":prometeu-compiler:prometeu-compiler-core"))
     implementation(project(":prometeu-compiler:prometeu-frontend-api"))
     implementation(project(":prometeu-compiler:prometeu-frontend-registry"))
-}
\ No newline at end of file
+}
+
+tasks.register<JavaExec>("runCompile") {
+    group = "application"
+    description = "Runs p.studio.compiler.Compile against test-projects/main."
+    classpath = sourceSets.main.get().runtimeClasspath
+    mainClass = "p.studio.compiler.Compile"
+    workingDir = rootDir
+}
diff --git a/test-projects/main/cartridge/assets.pa b/test-projects/main/cartridge/assets.pa
index e69de29b..b29c05c9 100644
Binary files a/test-projects/main/cartridge/assets.pa and b/test-projects/main/cartridge/assets.pa differ
diff --git a/test-projects/main/cartridge/manifest.json b/test-projects/main/cartridge/manifest.json
index 4f93d84a..901fe9e6 100644
--- a/test-projects/main/cartridge/manifest.json
+++ b/test-projects/main/cartridge/manifest.json
@@ -6,5 +6,5 @@
   "app_version": "0.1.0",
   "app_mode": "Game",
   "entrypoint": "0",
-  "capabilities": ["log", "gfx"]
+  "capabilities": ["log", "gfx", "asset"]
 }
diff --git a/test-projects/main/cartridge/program.pbx b/test-projects/main/cartridge/program.pbx
index a090a149..222af36f 100644
Binary files a/test-projects/main/cartridge/program.pbx and b/test-projects/main/cartridge/program.pbx differ
diff --git a/test-projects/main/run.sh b/test-projects/main/run.sh
index f840678a..3f792c15 100755
--- a/test-projects/main/run.sh
+++ b/test-projects/main/run.sh
@@ -2,5 +2,6 @@
 
 set -e
 
+cp build/assets.pa cartridge
 cp build/program.pbx cartridge
-./runtime/prometeu run cartridge
\ No newline at end of file
+./runtime/prometeu run cartridge
diff --git a/test-projects/main/src/main.pbs b/test-projects/main/src/main.pbs
index cd6c7fb9..f39b9981 100644
--- a/test-projects/main/src/main.pbs
+++ b/test-projects/main/src/main.pbs
@@ -9,7 +9,8 @@ fn frame() -> void
     let touch = Input.touch();
 
     Gfx.clear(new Color(6577));
-    Gfx.draw_square(touch.x() - 8, touch.y() - 8, 16, 16, new Color(65535), new Color(13271));
+    // Gfx.draw_square(touch.x() - 8, touch.y() - 8, 16, 16, new Color(65535), new Color(13271));
+    let sprite_status = Gfx.set_sprite(0, 0, touch.x() - 8, touch.y() - 8, 0, 1, true, false, false, 0);
 
     let a = 10;
     let b = 15;