Reply-length checker tests stop reaching the real judge

engine/hooks/diu-stop/tests/test_fix_matrix.py

@@ -3,11 +3,14 @@
 import unittest
 
 HOOKS_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+LLM_JUDGE_DIR = os.path.join(os.path.dirname(HOOKS_DIR), "llm-judge")
+sys.path.insert(0, LLM_JUDGE_DIR)
 sys.path.insert(0, HOOKS_DIR)
 sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
 
 import claude_stop_check  # noqa: E402
 from test_hooks import run_claude_check  # noqa: E402
+from judge_test_base import JudgeTestCase  # noqa: E402
 
 LONG_FILLER = " ".join(["word"] * (claude_stop_check.WORD_LIMIT + 20))
 
@@ -69,7 +72,7 @@
 ]
 
 
-class TestFixDoesNotTripAnotherCheck(unittest.TestCase):
+class TestFixDoesNotTripAnotherCheck(JudgeTestCase):
     """For each known trigger, the message a compliant rewrite would
     produce must not itself get blocked by any check -- otherwise fixing
     one finding just bounces you into another before the same-turn retry
@@ -91,7 +94,7 @@ def test_fixed_fixtures_are_not_blocked_by_any_check(self):
                 )
 
 
-class TestKnownDoubleBlocksResolveByNamingTheBlocker(unittest.TestCase):
+class TestKnownDoubleBlocksResolveByNamingTheBlocker(JudgeTestCase):
     """Some fixes deliberately still trip a second check (see
     KNOWN_DOUBLE_BLOCKS). `stop_hook_active` no longer releases those -- it
     only stops the word-count check, so a rewrite cannot smuggle a new

engine/hooks/diu-stop/tests/test_hooks.py

@@ -21,13 +21,16 @@
 
 HOOKS_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
 FIXTURES_DIR = os.path.join(os.path.dirname(os.path.abspath(__file__)), "fixtures")
+LLM_JUDGE_DIR = os.path.join(os.path.dirname(HOOKS_DIR), "llm-judge")
+sys.path.insert(0, LLM_JUDGE_DIR)
 sys.path.insert(0, HOOKS_DIR)
 
 import claude_prompt_reminder  # noqa: E402
 import claude_stop_check  # noqa: E402
 import codex_notify  # noqa: E402
 import install_claude_hook  # noqa: E402
 import install_codex_notify  # noqa: E402
+from judge_test_base import JudgeTestCase  # noqa: E402
 
 
 def run_claude_check(stdin_obj):
@@ -65,7 +68,7 @@ def run_codex_notify(argv_tail):
     return buf.getvalue()
 
 
-class TestClaudeStopCheck(unittest.TestCase):
+class TestClaudeStopCheck(JudgeTestCase):
     def test_under_limit_prints_nothing(self):
         blocked, err = run_claude_check({"last_assistant_message": "short reply"})
         self.assertFalse(blocked)
@@ -177,7 +180,7 @@ def test_malformed_stdin_json_does_not_crash(self):
         self.assertEqual(buf.getvalue(), "")
 
 
-class TestUnverifiedClaimCheck(unittest.TestCase):
+class TestUnverifiedClaimCheck(JudgeTestCase):
     """Regression tests for the three real unverified claims a session let
     through before self-correcting or being corrected by the user (see
     module docstring). Each `reproduces_the_incident` test asserts the OLD

engine/hooks/diu-stop/tests/test_limit_agreement.py

@@ -4,9 +4,14 @@
 import sys
 import unittest
 
-sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+TESTS_DIR = os.path.dirname(os.path.abspath(__file__))
+HOOKS_DIR = os.path.dirname(TESTS_DIR)
+LLM_JUDGE_DIR = os.path.join(os.path.dirname(HOOKS_DIR), "llm-judge")
+sys.path.insert(0, TESTS_DIR)
+sys.path.insert(0, LLM_JUDGE_DIR)
 
 from test_hooks import run_claude_check, run_prompt_reminder  # noqa: E402
+from judge_test_base import JudgeTestCase  # noqa: E402
 
 WORD_COUNT_RE = re.compile(r"\b(\d+) words\b")
 
@@ -28,7 +33,7 @@ def prose(n):
     return " ".join(["word"] * n)
 
 
-class TestReminderAndCheckerAgree(unittest.TestCase):
+class TestReminderAndCheckerAgree(JudgeTestCase):
     def test_reminder_states_exactly_one_word_limit(self):
         self.assertEqual(len(WORD_COUNT_RE.findall(reminder_text())), 1, reminder_text())
 

engine/hooks/diu-stop/tests/test_plain_words_check.py

@@ -8,9 +8,12 @@
 from unittest.mock import patch
 
 HOOK_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+LLM_JUDGE_DIR = os.path.join(os.path.dirname(HOOK_DIR), "llm-judge")
+sys.path.insert(0, LLM_JUDGE_DIR)
 sys.path.insert(0, HOOK_DIR)
 
 import plain_words
+from judge_test_base import JudgeTestCase
 
 REPLY = "No hook decides differently. Preflight passes and the review unit is engine-runtime."
 ASKED = "Is it safe?"
@@ -36,14 +39,16 @@ def _hit(job_id, closest, category="plain-words-made-up-labels"):
     return {"id": job_id, "hook": plain_words.HOOK_NAME, "outcome": "hit", "answer": {"match": True, "category": category, "closest": closest}}
 
 
-class PlainWordsCase(unittest.TestCase):
+class PlainWordsCase(JudgeTestCase):
     def setUp(self):
+        super().setUp()
         self._tmp = tempfile.TemporaryDirectory()
         self.tmp = self._tmp.name
         self.judge, _ = plain_words._llm_judge()
 
     def tearDown(self):
         self._tmp.cleanup()
+        super().tearDown()
 
     def payload(self, rows=(_user(ASKED), _assistant(REPLY)), **extra):
         payload = {"last_assistant_message": REPLY}