DaoyuanLi2816 · DaoyuanLi2816 · Aug 6, 2026 · Aug 6, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -6,6 +6,15 @@ All notable changes to miniVERL are recorded here. The format follows
 
 ## [Unreleased]
 
+### Fixed
+
+- A reward scaffold saved with a UTF-8 byte-order mark is no longer reported as
+  a syntax error. CPython strips the BOM when it reads a source file, so such a
+  scaffold imports normally; the static checker handed the leading `U+FEFF` to
+  `ast.parse`, which rejected it. The failure was fail-closed — a legitimate
+  file was refused, nothing unsafe was accepted — and a BOM still cannot hide a
+  top-level call. Found while verifying the published v0.6.3 wheel on Windows.
+
 ## [0.6.3] - 2026-08-05
 
 ### Security

diff --git a/docs/release-checklist.md b/docs/release-checklist.md
@@ -6,11 +6,17 @@ after the exact release commit and its remote checks are green.
 
 ## v0.6.4 (in development)
 
-- [ ] A reward scaffold saved with a UTF-8 byte-order mark parses instead of
+- [x] A reward scaffold saved with a UTF-8 byte-order mark parses instead of
       being reported as a syntax error. CPython strips the BOM when it reads a
       source file; the static checker did not, so a scaffold written by a
       Windows editor was refused as unparseable. Found while verifying the
-      published v0.6.3 wheel.
+      published v0.6.3 wheel. Two regressions cover it, including one proving a
+      BOM cannot hide a top-level call.
+- [ ] `pinned-profile-smoke` remains excluded from the required-status-check
+      list because `verl-bridge.yml` filters on paths and therefore never
+      reports on an unrelated pull request. It still runs, and must pass, on any
+      pull request that touches bridge code. Revisit if the workflow ever loses
+      its path filter.
 
 ## v0.6.3 Security, artifact integrity and release-state hardening
 

diff --git a/src/miniverl/bridge/reward_static.py b/src/miniverl/bridge/reward_static.py
@@ -526,6 +526,10 @@ def _finish(findings: list[dict[str, Any]]) -> dict[str, Any]:
     except UnicodeDecodeError as exc:
         check["detail"] = f"reward scaffold is not valid UTF-8: {exc.reason}"
         return _finish([_finding("encoding_error", 0, "source is not decodable as UTF-8")])
+    # CPython strips a UTF-8 BOM when it reads a source file, so a scaffold
+    # saved by a Windows editor imports fine. Handing the leading U+FEFF to
+    # ast.parse instead reports a perfectly good file as a syntax error.
+    source = source.lstrip("")
 
     try:
         tree = ast.parse(source, filename=str(source_path))

diff --git a/tests/unit/test_verl_bridge_reward_definition_time.py b/tests/unit/test_verl_bridge_reward_definition_time.py
@@ -218,6 +218,35 @@ def test_findings_are_bounded(tmp_path: Path) -> None:
     assert check["findings_truncated"] is True
 
 
+def test_a_byte_order_mark_does_not_look_like_a_syntax_error(tmp_path: Path) -> None:
+    """Found while verifying the published v0.6.3 wheel on Windows.
+
+    CPython strips a UTF-8 BOM when it reads a source file, so this scaffold
+    imports fine; handing the leading U+FEFF to ``ast.parse`` reported it as
+    unparseable instead.
+    """
+    target = tmp_path / "reward_or_verifier_scaffold.py"
+    target.write_bytes(b"\xef\xbb\xbf" + VALID.encode("utf-8"))
+
+    check = inspect_reward_scaffold(target)
+
+    assert check["status"] == "ok"
+    assert check["verification_level"] == "interface_shape_verified"
+    assert check["findings"] == []
+
+
+def test_a_byte_order_mark_does_not_hide_a_top_level_call(tmp_path: Path) -> None:
+    target = tmp_path / "reward_or_verifier_scaffold.py"
+    target.write_bytes(b"\xef\xbb\xbf" + (PRELUDE + "exploit()\n\n\n" + VALID).encode("utf-8"))
+
+    check = inspect_reward_scaffold(target)
+
+    assert check["status"] == "fail"
+    assert "top_level_call" in _categories(check)
+    assert check["code_executed"] is False
+    assert not (tmp_path / "PWNED.txt").exists()
+
+
 def test_undecodable_source_fails_closed(tmp_path: Path) -> None:
     target = tmp_path / "reward_or_verifier_scaffold.py"
     target.write_bytes(b"\xff\xfe\x00invalid utf-8 \xc3\x28\n")