Clarify Harbor smoke verifier scores (part 4)

harbor/tasks/v3_depth3_net_revenue_usd_t14/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t15/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t16/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t17/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t18/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t19/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t2/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t20/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t21/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t22/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t23/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t24/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t3/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t4/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t5/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t6/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t7/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t8/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_net_revenue_usd_t9/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t0/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t1/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t10/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t11/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t12/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t13/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t14/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t15/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t16/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t17/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t18/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t19/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t2/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t20/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t21/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t22/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1
 
 

harbor/tasks/v3_depth3_refund_share_bp_t23/tests/check_output_shape.py

@@ -7,12 +7,20 @@ import re
 from pathlib import Path
 
 
-def write_reward(score: float, diagnostic: str) -> None:
+def write_reward(passed_smoke: bool, diagnostic: str) -> None:
     log_dir = Path(os.environ.get("LOG_DIR", "/logs/verifier"))
     log_dir.mkdir(parents=True, exist_ok=True)
-    (log_dir / "reward.txt").write_text(f"{score:.6g}\n")
+    (log_dir / "reward.txt").write_text("0\n")
     (log_dir / "reward.json").write_text(
-        json.dumps({"reward": score, "smoke_only": True}, indent=2) + "\n"
+        json.dumps(
+            {
+                "reward": 0.0,
+                "passed_smoke": passed_smoke,
+                "official_score": False,
+                "smoke_only": True,
+            },
+            indent=2,
+        ) + "\n"
     )
     (log_dir / "diagnostic.txt").write_text(diagnostic + "\n")
 
@@ -22,7 +30,7 @@ def main() -> int:
     task = json.loads((app_dir / "task.json").read_text())
     answer_path = app_dir / "answer.txt"
     if not answer_path.exists():
-        write_reward(0.0, "missing /app/answer.txt")
+        write_reward(False, "missing /app/answer.txt")
         return 1
 
     got = answer_path.read_text().strip()
@@ -34,7 +42,7 @@ def main() -> int:
         diagnostic = "valid code smoke output" if ok else f"not an exact code: {got!r}"
 
     # This is intentionally not official scoring; it only verifies I/O shape.
-    write_reward(1.0 if ok else 0.0, diagnostic)
+    write_reward(ok, diagnostic)
     return 0 if ok else 1