audit(token_audit): count "bogus" and "it didn't run" as the same class

engine/skills/reflect/scripts/tests/test_token_audit.py

@@ -658,7 +658,7 @@ def test_product_blame_and_plain_questions_do_not_flag_intervention(self):
         lines = [
             codex_response_item("user", "the UI is messed up on the workers page", ts="2026-09-09T04:40:47.000Z"),
             codex_response_item("assistant", "Looking at the panel now.", ts="2026-09-09T04:41:00.000Z"),
-            codex_response_item("user", "what would you recommend for the deploy window?", ts="2026-09-09T05:46:08.000Z"),
+            codex_response_item("user", "why did you choose Saturday for the deploy window?", ts="2026-09-09T05:46:08.000Z"),
             codex_response_item("assistant", "Three options below.", ts="2026-09-09T05:47:00.000Z"),
         ]
         path = write_jsonl(lines)
@@ -670,6 +670,26 @@ def test_product_blame_and_plain_questions_do_not_flag_intervention(self):
         finally:
             os.unlink(path)
 
+    def test_handed_over_something_broken_flags_intervention(self):
+        """The wording that scored zero: no profanity, no told-you. The user
+        says the artifact was bogus, then that it did not run."""
+        lines = [
+            codex_response_item("user", "huh? why did you give me a bogus script?", ts="2026-09-11T05:40:00.000Z"),
+            codex_response_item("assistant", "Fixing the quoting now.", ts="2026-09-11T05:41:00.000Z"),
+            codex_response_item("user", "it still didn't run", ts="2026-09-11T05:50:00.000Z"),
+            codex_response_item("assistant", "Smoke-testing the transport first.", ts="2026-09-11T05:51:00.000Z"),
+        ]
+        path = write_jsonl(lines)
+        try:
+            with redirect_stdout(io.StringIO()):
+                result = token_audit.audit_codex(path)
+            kinds = {k for f in result["frustration"]["flagged"] for k in f["kinds"]}
+            self.assertIn("cheap-way-out", kinds)
+            flags = {fl["name"]: fl for fl in result["flags"]}
+            self.assertEqual(flags["intervention-must-automate"]["value"], "yes")
+        finally:
+            os.unlink(path)
+
     def test_committed_codex_fixture_hits_frustration_and_intervention(self):
         result = token_audit.audit_codex(fixture("codex_thrash_session.jsonl"))
         flags = {fl["name"]: fl for fl in result["flags"]}

engine/skills/reflect/scripts/token_audit.py

@@ -142,7 +142,7 @@ def _direct_run_targets(command):
         r"|\bis the proof att?ach?ed\b"
         r"|\bwhere'?s the proof\b"
         r"|\bprove (to me )?that (it|this|that|the|#?\d)", re.I)),
-    ("cheap-way-out", re.compile(r"\bcheap way out\b|\bwhy would you\b|\bthat'?s (weird|wierd)\b|\bstraight up\b", re.I)),
+    ("cheap-way-out", re.compile(r"\bcheap way out\b|\bwhy would you\b|\bbogus\b|\bdidn'?t (even )?(work|run)\b|\bthat'?s (weird|wierd)\b|\bstraight up\b", re.I)),
     ("explicit-invocation", re.compile(r"(?:^|\s)/(?:automate-me|reflect|thrash)\b", re.I)),
 ]