diff --git a/scripts/ci/opencode_review_normalize_output.py b/scripts/ci/opencode_review_normalize_output.py
index f16ed6ec9..1f20a2084 100755
--- a/scripts/ci/opencode_review_normalize_output.py
+++ b/scripts/ci/opencode_review_normalize_output.py
@@ -864,6 +864,31 @@ def reject(reason: str) -> int:
     return 0
 
 
+def canonicalize_finding_fields(finding: dict[str, Any]) -> dict[str, Any]:
+    """Map known-safe model vocabulary drift onto the canonical finding schema.
+
+    Findings only exist on REQUEST_CHANGES control blocks (valid_control rejects
+    APPROVE blocks that carry findings), so rescuing a drifted finding can only
+    publish a blocking review — it can never loosen approval evidence. The
+    observed safe drift is repaired: ``priority`` used in place of
+    ``severity``. Source-backed ``suggested_diff`` evidence must remain
+    explicit because the downstream publication gate verifies it against the
+    current-head diff.
+    """
+
+    def has_non_blank_text(field_candidate: Any) -> bool:
+        """Return whether a field candidate is a non-blank string."""
+        return isinstance(field_candidate, str) and bool(field_candidate.strip())
+
+    finding = dict(finding)
+    priority = finding.pop("priority", None)
+    if not has_non_blank_text(finding.get("severity")) and has_non_blank_text(
+        priority
+    ):
+        finding["severity"] = priority
+    return finding
+
+
 def valid_control(
     value: Any,
     *,
@@ -944,15 +969,18 @@ def valid_control(
         "regression_test_direction",
         "suggested_diff",
     )
+    normalized_findings = []
     for finding in findings:
         if not isinstance(finding, dict):
             return None
         line = finding.get("line")
         if isinstance(line, bool) or not isinstance(line, int) or line <= 0:
             return None
+        finding = canonicalize_finding_fields(finding)
         for field in required_finding_fields:
             if not isinstance(finding.get(field), str) or not finding[field].strip():
                 return None
+        normalized_findings.append(finding)
 
     normalized = {
         "head_sha": value["head_sha"],
@@ -961,7 +989,7 @@ def valid_control(
         "result": result,
         "reason": reason,
         "summary": summary,
-        "findings": findings,
+        "findings": normalized_findings,
     }
     if isinstance(value.get("adversarial_validation"), dict):
         normalized["adversarial_validation"] = value["adversarial_validation"]
diff --git a/tests/test_opencode_review_normalize_output.py b/tests/test_opencode_review_normalize_output.py
index 2c817b63d..5b48f22f0 100644
--- a/tests/test_opencode_review_normalize_output.py
+++ b/tests/test_opencode_review_normalize_output.py
@@ -1,4 +1,7 @@
 import json
+import shutil
+import subprocess
+from pathlib import Path
 
 import pytest
 
@@ -725,6 +728,126 @@ def test_valid_control_filters_shape_head_and_review_contract():
     assert norm.valid_control(approve_without_findings_key, **kwargs)["findings"] == []
 
 
+def test_valid_control_canonicalizes_known_safe_finding_field_drift():
+    kwargs = {
+        "expected_head_sha": "head",
+        "expected_run_id": "run",
+        "expected_run_attempt": "attempt",
+    }
+
+    aliased = finding(priority="P1")
+    del aliased["severity"]
+    normalized = norm.valid_control(
+        control(result="REQUEST_CHANGES", findings=[aliased]), **kwargs
+    )
+    assert normalized is not None
+    assert normalized["findings"][0]["severity"] == "P1"
+    assert "priority" not in normalized["findings"][0]
+
+    diffless = finding()
+    del diffless["suggested_diff"]
+    assert (
+        norm.valid_control(
+            control(result="REQUEST_CHANGES", findings=[diffless]), **kwargs
+        )
+        is None
+    )
+
+    blank_diff = finding(suggested_diff="  ")
+    assert (
+        norm.valid_control(
+            control(result="REQUEST_CHANGES", findings=[blank_diff]), **kwargs
+        )
+        is None
+    )
+
+    canonical_severity_wins = finding(priority="P2")
+    normalized = norm.valid_control(
+        control(result="REQUEST_CHANGES", findings=[canonical_severity_wins]),
+        **kwargs,
+    )
+    assert normalized is not None
+    assert normalized["findings"][0]["severity"] == "HIGH"
+    assert "priority" not in normalized["findings"][0]
+
+    blank_alias = finding(priority="   ")
+    del blank_alias["severity"]
+    assert (
+        norm.valid_control(
+            control(result="REQUEST_CHANGES", findings=[blank_alias]), **kwargs
+        )
+        is None
+    )
+
+    no_remedy = finding(fix_direction="", suggested_diff="")
+    assert (
+        norm.valid_control(
+            control(result="REQUEST_CHANGES", findings=[no_remedy]), **kwargs
+        )
+        is None
+    )
+
+
+def test_approval_gate_rejects_prose_fix_direction_without_suggested_diff(tmp_path):
+    bash_command = shutil.which("bash")
+    if bash_command is None:
+        pytest.skip("bash is unavailable")
+    try:
+        subprocess.run(
+            [bash_command, "--version"],
+            capture_output=True,
+            text=True,
+            timeout=5,
+            check=True,
+        )
+    except (OSError, subprocess.SubprocessError) as exc:
+        pytest.skip(f"bash is not usable for this regression test: {exc}")
+
+    repo_root = Path(__file__).resolve().parents[1]
+    gate_script = repo_root / "scripts" / "ci" / "opencode_review_approve_gate.sh"
+    control_data = control(
+        result="REQUEST_CHANGES",
+        findings=[finding(fix_direction="Restore the guard.")],
+    )
+    del control_data["findings"][0]["suggested_diff"]
+    comment_file = tmp_path / "comment.md"
+    comment_file.write_text(
+        "\n".join(
+            [
+                "<!-- opencode-review-gate head_sha=head run_id=run run_attempt=attempt -->",
+                "<!-- opencode-review-control-v1",
+                json.dumps(control_data),
+                "-->",
+                "",
+            ]
+        ),
+        encoding="utf-8",
+    )
+
+    completed_process = subprocess.run(
+        [
+            bash_command,
+            str(gate_script),
+            "head",
+            "run",
+            "attempt",
+            str(comment_file),
+        ],
+        cwd=repo_root,
+        capture_output=True,
+        text=True,
+        timeout=20,
+        check=False,
+    )
+
+    assert completed_process.returncode == 4
+    assert completed_process.stdout.strip() == "NO_CONCLUSION"
+    assert (
+        "finding 1 field suggested_diff must be a non-empty string"
+        in completed_process.stderr
+    )
+
+
 def test_valid_control_repairs_approval_summary_from_bounded_evidence(tmp_path, monkeypatch):
     evidence = tmp_path / "bounded-review-evidence.md"
     evidence.write_text(