diff --git a/CHANGELOG.md b/CHANGELOG.md
index 8673520..5f86489 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -12,6 +12,9 @@ All notable changes to SkillEvaluator are documented in this file.
 
 ### Fixed
 
+- Malformed, non-UTF-8, or unreadable bundled and custom policy files now
+  produce path-specific CLI errors instead of leaking raw parser or I/O errors
+  ([#128](https://github.com/NVIDIA/SkillEvaluator/issues/128)).
 - License detection no longer treats a frontmatter `license` identifier as
   authoritative when a LICENSE file declares a different license. Claiming
   MIT while shipping GPL-3.0 now fails closed. Every LICENSE/COPYING file is
diff --git a/src/skillevaluator/validators/policy.py b/src/skillevaluator/validators/policy.py
index e363435..89a09e9 100644
--- a/src/skillevaluator/validators/policy.py
+++ b/src/skillevaluator/validators/policy.py
@@ -161,6 +161,19 @@ def _coerce_email_regex(value: Any, source: str) -> re.Pattern[str] | None:
 _KNOWN_IDENTITY_KEYS = {"author_email_regex"}
 
 
+def _load_policy_yaml(path: Path) -> Any:
+    """Load policy YAML and normalize parser failures to the public contract."""
+    try:
+        with path.open(encoding="utf-8") as fh:
+            return yaml.safe_load(fh) or {}
+    except (yaml.YAMLError, UnicodeError, RecursionError) as exc:
+        raise ValueError(f"Invalid policy YAML in {path}: {exc}") from exc
+    except FileNotFoundError:
+        raise
+    except OSError as exc:
+        raise ValueError(f"Could not read policy file {path}: {exc}") from exc
+
+
 def _warn_unknown_keys(data: dict[str, Any], known: set[str], context: str, source: str) -> None:
     unknown = set(data) - known
     if unknown:
@@ -227,8 +240,7 @@ def load_profile(name: str = DEFAULT_PROFILE_NAME) -> ValidationPolicy:
     if not path.exists():
         available = sorted(p.stem for p in PROFILES_DIR.glob("*.yaml")) if PROFILES_DIR.exists() else []
         raise FileNotFoundError(f"Unknown profile {name!r}. Available bundled profiles: {available or '(none)'}")
-    with path.open(encoding="utf-8") as fh:
-        data = yaml.safe_load(fh) or {}
+    data = _load_policy_yaml(path)
     return _policy_from_data(data, fallback_profile=name, source=path)
 
 
@@ -248,8 +260,7 @@ def load_policy_file(
         raise FileNotFoundError(f"Custom policy file not found: {path}")
 
     base = load_profile(base_profile)
-    with path.open(encoding="utf-8") as fh:
-        custom_data = yaml.safe_load(fh) or {}
+    custom_data = _load_policy_yaml(path)
     custom = _policy_from_data(custom_data, fallback_profile=base.profile, source=path)
 
     merged_overrides = dict(base.severity_overrides)
diff --git a/tests/test_commands.py b/tests/test_commands.py
index b9e28b5..a91abe0 100644
--- a/tests/test_commands.py
+++ b/tests/test_commands.py
@@ -93,6 +93,29 @@ def test_validate_fixture_no_llm() -> None:
     assert "All validations passed" in result.output
 
 
+def test_validate_reports_malformed_policy_without_traceback(tmp_path: Path) -> None:
+    policy = tmp_path / "broken-policy.yaml"
+    policy.write_text("severity_overrides: [", encoding="utf-8")
+
+    result = CliRunner().invoke(
+        cli,
+        [
+            "validate",
+            str(FIXTURE),
+            "--no-llm",
+            "--no-dedup",
+            "--checks",
+            "schema",
+            "--policy",
+            str(policy),
+        ],
+    )
+
+    assert result.exit_code == 1
+    assert f"Invalid policy YAML in {policy}" in result.output
+    assert isinstance(result.exception, SystemExit)
+
+
 def test_validate_prints_tier1_section_banner() -> None:
     # The Tier 1 section is announced as it runs so it is visibly reported in
     # CI logs (SkillEvaluator parity), not only inside the final combined report.
diff --git a/tests/validators/test_policy.py b/tests/validators/test_policy.py
index ccfe701..f6bffba 100644
--- a/tests/validators/test_policy.py
+++ b/tests/validators/test_policy.py
@@ -8,8 +8,10 @@
 from pathlib import Path
 
 import pytest
+import yaml
 
 from skillevaluator.models.result import Severity
+from skillevaluator.validators import policy as policy_module
 from skillevaluator.validators.policy import (
     DEFAULT_PROFILE_NAME,
     ValidationPolicy,
@@ -46,6 +48,75 @@ def test_custom_policy_overlays_the_public_profile(tmp_path: Path) -> None:
     assert policy.severity_for("SCHEMA", "author_missing", Severity.LOW) == Severity.HIGH
 
 
+def test_custom_policy_wraps_malformed_yaml(tmp_path: Path) -> None:
+    custom = tmp_path / "broken-policy.yaml"
+    custom.write_text("severity_overrides: [", encoding="utf-8")
+
+    with pytest.raises(ValueError) as exc_info:
+        load_policy_file(custom)
+
+    assert f"Invalid policy YAML in {custom}" in str(exc_info.value)
+    assert isinstance(exc_info.value.__cause__, yaml.YAMLError)
+
+
+def test_custom_policy_wraps_invalid_encoding(tmp_path: Path) -> None:
+    custom = tmp_path / "utf16-policy.yaml"
+    custom.write_bytes("profile: encoded\n".encode("utf-16"))
+
+    with pytest.raises(ValueError) as exc_info:
+        load_policy_file(custom)
+
+    assert f"Invalid policy YAML in {custom}" in str(exc_info.value)
+    assert isinstance(exc_info.value.__cause__, UnicodeDecodeError)
+
+
+def test_custom_policy_wraps_read_errors(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
+    custom = tmp_path / "unreadable-policy.yaml"
+    custom.write_text("profile: unreadable\n", encoding="utf-8")
+    original_open = Path.open
+
+    def deny_custom_policy(path: Path, *args: object, **kwargs: object):
+        if path == custom:
+            raise PermissionError("access denied")
+        return original_open(path, *args, **kwargs)
+
+    monkeypatch.setattr(Path, "open", deny_custom_policy)
+
+    with pytest.raises(ValueError) as exc_info:
+        load_policy_file(custom)
+
+    assert f"Could not read policy file {custom}" in str(exc_info.value)
+    assert isinstance(exc_info.value.__cause__, PermissionError)
+
+
+def test_policy_yaml_wraps_recursion_errors(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
+    policy = tmp_path / "deeply-nested-policy.yaml"
+    policy.write_text("severity_overrides: []\n", encoding="utf-8")
+
+    def exhaust_parser(_stream: object) -> object:
+        raise RecursionError("maximum recursion depth exceeded")
+
+    monkeypatch.setattr(policy_module.yaml, "safe_load", exhaust_parser)
+
+    with pytest.raises(ValueError) as exc_info:
+        policy_module._load_policy_yaml(policy)
+
+    assert f"Invalid policy YAML in {policy}" in str(exc_info.value)
+    assert isinstance(exc_info.value.__cause__, RecursionError)
+
+
+def test_bundled_profile_wraps_malformed_yaml(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
+    profile = tmp_path / "broken.yaml"
+    profile.write_text("identity: [", encoding="utf-8")
+    monkeypatch.setattr(policy_module, "PROFILES_DIR", tmp_path)
+
+    with pytest.raises(ValueError) as exc_info:
+        load_profile("broken")
+
+    assert f"Invalid policy YAML in {profile}" in str(exc_info.value)
+    assert isinstance(exc_info.value.__cause__, yaml.YAMLError)
+
+
 def test_policy_validation_and_resolution() -> None:
     assert resolve_policy().profile == "external"
     assert (