fix(tests): correct Cyrillic-A assertion to expect invalid-character warning

NFKC normalization does not map Cyrillic lookalikes to Latin characters
(e.g. Cyrillic А U+0410 normalises to itself, not to Latin A U+0041).
These characters are rejected correctly by the existing ASCII-only regex
check, not by the new NFKC check. The test was asserting the wrong warning
key; update it to assert the "invalid characters" diagnostic that is
actually emitted, and add a comment explaining the distinction.

Fixes review finding from cubic-dev-ai.

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>

Author: manjunathbhaskar

tests/shared/test_tool_name_validation.py

@@ -225,13 +225,17 @@ class TestUnicodeHomoglyphDetection:
     """Tests for detecting Unicode homoglyphs that visually impersonate ASCII."""
 
     def test_cyrillic_a_rejected(self) -> None:
-        """Cyrillic А (U+0410) looks identical to Latin A but should be detected."""
+        """Cyrillic А (U+0410) looks identical to Latin A but must be rejected.
+
+        Note: NFKC does not map Cyrillic to Latin lookalikes, so this character
+        is caught by the ASCII-only regex check (invalid characters), not by the
+        normalization check. The rejection is the important invariant here.
+        """
         result = validate_tool_name("tool_А")
         assert not result.is_valid, "Cyrillic А should be rejected"
         assert any(
-            "normalize" in w.lower() or "homoglyph" in w.lower()
-            for w in result.warnings
-        ), f"Warning should mention normalization or homoglyph. Got: {result.warnings}"
+            "invalid characters" in w.lower() for w in result.warnings
+        ), f"Expected invalid-character warning. Got: {result.warnings}"
 
     def test_cyrillic_o_rejected(self) -> None:
         """Cyrillic О (U+041E) looks identical to Latin O."""