[eval] Cap TaskTrove verifiers at ten minutes

hpc/harbor_yaml/eval/snowball_tasktrove_v342_terminus2.yaml

@@ -28,8 +28,8 @@ environment:
   kwargs:
     auto_snapshot: true
 verifier:
-  override_timeout_sec: 14400
-  max_timeout_sec: 14400
+  override_timeout_sec: null
+  max_timeout_sec: 600
 agents:
   - name: terminus-2
     model_name: placeholder/override-at-runtime

hpc/harbor_yaml/eval/tasktrove_pi_ctx32k.yaml

@@ -23,8 +23,8 @@ environment:
   kwargs:
     auto_snapshot: true
 verifier:
-  override_timeout_sec: 7200
-  max_timeout_sec: 7200
+  override_timeout_sec: null
+  max_timeout_sec: 600
 agents:
   - name: pi
     model_name: placeholder/override-at-runtime

hpc/harbor_yaml/eval/tasktrove_qwen3_coder_ctx32k.yaml

@@ -23,7 +23,7 @@ environment:
   kwargs:
     auto_snapshot: true
 verifier:
-  max_timeout_sec: 14400
+  max_timeout_sec: 600
 agents:
   - name: terminus-2
     model_name: placeholder/override-at-runtime

hpc/harbor_yaml/eval/tasktrove_terminus2_ctx32k.yaml

@@ -23,7 +23,7 @@ environment:
   kwargs:
     auto_snapshot: true
 verifier:
-  max_timeout_sec: 14400
+  max_timeout_sec: 600
 agents:
   - name: terminus-2
     model_name: placeholder/override-at-runtime

hpc/harbor_yaml/eval/tasktrove_v44_glm52_terminus2_ctx32k.yaml

@@ -26,8 +26,8 @@ environment:
   kwargs:
     auto_snapshot: true
 verifier:
-  override_timeout_sec: 14400
-  max_timeout_sec: 14400
+  override_timeout_sec: null
+  max_timeout_sec: 600
 agents:
   - name: terminus-2
     model_name: placeholder/override-at-runtime

hpc/harbor_yaml/eval/tasktrove_v44_qwen35_opencode_ctx131k.yaml

@@ -26,8 +26,8 @@ environment:
   kwargs:
     auto_snapshot: true
 verifier:
-  override_timeout_sec: 14400
-  max_timeout_sec: 14400
+  override_timeout_sec: null
+  max_timeout_sec: 600
 agents:
   - name: opencode
     model_name: openai/override-at-runtime

hpc/harbor_yaml/eval/tasktrove_v46_qwen3_coder_terminus2_ctx32k.yaml

@@ -27,8 +27,8 @@ environment:
   kwargs:
     auto_snapshot: true
 verifier:
-  override_timeout_sec: 14400
-  max_timeout_sec: 14400
+  override_timeout_sec: null
+  max_timeout_sec: 600
 agents:
   - name: terminus-2
     model_name: placeholder/override-at-runtime

tests/hpc/test_tasktrove_eval_configs.py

@@ -0,0 +1,29 @@
+from pathlib import Path
+
+import yaml
+
+
+REPOSITORY_ROOT = Path(__file__).resolve().parents[2]
+EVAL_CONFIG_DIR = REPOSITORY_ROOT / "hpc/harbor_yaml/eval"
+MAX_VERIFIER_TIMEOUT = 600
+
+
+def tasktrove_eval_configs() -> list[Path]:
+    return sorted(
+        {
+            *EVAL_CONFIG_DIR.glob("tasktrove*.yaml"),
+            *EVAL_CONFIG_DIR.glob("snowball_tasktrove*.yaml"),
+        }
+    )
+
+
+def test_tasktrove_verifiers_are_capped_at_ten_minutes() -> None:
+    paths = tasktrove_eval_configs()
+    assert paths
+
+    for path in paths:
+        config = yaml.safe_load(path.read_text())
+        verifier = config["verifier"]
+        assert verifier["max_timeout_sec"] <= MAX_VERIFIER_TIMEOUT, path
+        override = verifier.get("override_timeout_sec")
+        assert override is None or override <= MAX_VERIFIER_TIMEOUT, path