diff --git a/CHANGELOG.md b/CHANGELOG.md
index 1cf2a082..5dc296a7 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -112,6 +112,8 @@
 
 * `senkin_tmp_train`, `senkin_tmp_predict`: Store the model bundle (the test predictions, as the solution is transductive) as an h5ad file in a directory instead of a pickle. The train image is built from `nvidia/cuda` with the current pandas 3, the predict image `openproblems/base_pytorch_nvidia:1` ships pandas 2, and unpickling the bundled `var` DataFrame raised `NotImplementedError` in `NDArrayBacked.__setstate__`, so senkin_tmp finished both CITE training runs of run_2026-09-23 but never produced a prediction. The predict step still reads the old pickle until the test resources are regenerated (PR #73).
 
+* `scipenn`: Exit 99 (non-applicable) instead of raising a `ValueError` on anything other than GEX -> ADT (PR #77).
+
 # task_predict_modality 0.1.1
 
 ## NEW FUNCTIONALITY
diff --git a/src/methods/scipenn/config.vsh.yaml b/src/methods/scipenn/config.vsh.yaml
index cd19413e..8b4d170e 100644
--- a/src/methods/scipenn/config.vsh.yaml
+++ b/src/methods/scipenn/config.vsh.yaml
@@ -44,6 +44,7 @@ arguments:
 resources:
   - type: python_script
     path: script.py
+  - path: /src/utils/exit_codes.py
 engines:
   # sciPENN is installed --no-deps: it pins a spurious numba<=0.50.0 (never imported) and a
   # 2021-era stack, but runs on the base image's torch/scanpy/anndata. The base already ships
diff --git a/src/methods/scipenn/script.py b/src/methods/scipenn/script.py
index 21a8e200..566acffd 100644
--- a/src/methods/scipenn/script.py
+++ b/src/methods/scipenn/script.py
@@ -1,5 +1,6 @@
 import logging
 import os
+import sys
 import tempfile
 
 import anndata as ad
@@ -23,9 +24,12 @@
     "batch_size": 128,
     "seed": 0,
 }
-meta = {"name": "scipenn"}
+meta = {"name": "scipenn", "resources_dir": "src/utils"}
 ## VIASH END
 
+sys.path.append(meta["resources_dir"])
+from exit_codes import exit_non_applicable
+
 logger.info("Reading input files...")
 rna_train = ad.read_h5ad(par["input_train_mod1"])   # mod1 = RNA (GEX)
 prot_train = ad.read_h5ad(par["input_train_mod2"])   # mod2 = protein (ADT)
@@ -33,11 +37,11 @@
 
 # sciPENN only supports the GEX -> ADT direction. It biologically interprets the
 # gene sets as RNA and the protein sets as surface protein, so running it on the
-# swapped (ADT -> GEX) direction produces meaningless output. Fail loudly instead.
+# swapped (ADT -> GEX) direction produces meaningless output.
 mod1 = rna_train.uns.get("modality")
 mod2 = prot_train.uns.get("modality")
 if mod1 != "GEX" or mod2 != "ADT":
-    raise ValueError(
+    exit_non_applicable(
         f"sciPENN only supports predicting protein (ADT) from RNA (GEX); "
         f"got mod1={mod1!r}, mod2={mod2!r}."
     )