From 99b6a9bb8f878f151dab7d3f996a4abfe4e62cb6 Mon Sep 17 00:00:00 2001
From: Eneko Ruiz <196690101+eneekoruiz@users.noreply.github.com>
Date: Tue, 13 Oct 2026 12:05:24 +0200
Subject: [PATCH] fix: reject lossy classifier tool-call indices

RED: 6 failures proved booleans and fractional indices could invent evidence. GREEN: 49 targeted tests; full skill-comply suites pass on Windows (88 passed, 5 platform skips) and Linux (93 passed). Parser coverage 100%; exact integer floats and numeric strings remain supported. Pi/catalog checks pass.
(cherry picked from commit 6dde588860b440313c3cf63561f10203fcf28f55)
---
 skills/skill-comply/scripts/classifier.py    | 2 ++
 skills/skill-comply/tests/test_classifier.py | 9 +++++++--
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/skills/skill-comply/scripts/classifier.py b/skills/skill-comply/scripts/classifier.py
index 76d7c87aa..74361c8bf 100644
--- a/skills/skill-comply/scripts/classifier.py
+++ b/skills/skill-comply/scripts/classifier.py
@@ -85,6 +85,8 @@ def _parse_classification(text: str) -> dict[str, list[int]]:
                 raise ClassifierParseError(
                     f"classifier returned non-list JSON value: {type(value).__name__}"
                 )
+            if any(isinstance(i, bool) or (isinstance(i, float) and not i.is_integer()) for i in value):
+                raise ValueError("tool call indices must be whole numbers, not booleans")
         return {k: [int(i) for i in v] for k, v in parsed.items()}
     except (json.JSONDecodeError, ValueError, TypeError, OverflowError) as e:
         raise ClassifierParseError(
diff --git a/skills/skill-comply/tests/test_classifier.py b/skills/skill-comply/tests/test_classifier.py
index a94f44fe5..a9e80a599 100644
--- a/skills/skill-comply/tests/test_classifier.py
+++ b/skills/skill-comply/tests/test_classifier.py
@@ -28,7 +28,12 @@ class TestParseClassification:
             _parse_classification('{"a": [0], "other": "x"}')
 
     def test_existing_index_coercion_is_preserved(self) -> None:
-        assert _parse_classification('{"a": [true, 2.9, "3"]}') == {"a": [1, 2, 3]}
+        assert _parse_classification('{"a": [0, 2.0, "3"]}') == {"a": [0, 2, 3]}
+
+    @pytest.mark.parametrize("index", [True, False, 2.9, -0.1])
+    def test_boolean_and_fractional_indices_raise(self, index: object) -> None:
+        with pytest.raises(classifier.ClassifierParseError, match="failed to parse"):
+            _parse_classification(json.dumps({"a": [index]}))
 
     def test_empty_object_is_valid(self) -> None:
         assert _parse_classification("{}") == {}
@@ -54,7 +59,7 @@ class TestParseClassification:
         assert issubclass(classifier.ClassifierParseError, RuntimeError)
 
 
-@pytest.mark.parametrize("response", ["not JSON", "[]", '{"a": ["x"]}', '{"a": "x"}', '{"a": [0], "other": null}', '{"a": [1e309]}', '{"a": [-1e309]}'])
+@pytest.mark.parametrize("response", ["not JSON", "[]", '{"a": ["x"]}', '{"a": "x"}', '{"a": [0], "other": null}', '{"a": [1e309]}', '{"a": [-1e309]}', '{"a": [false]}', '{"a": [0.5]}'])
 def test_grader_propagates_classifier_parse_failure(response: str) -> None:
     """An invalid successful subprocess response cannot become a zero score."""
     spec = ComplianceSpec(