diff --git a/.github/workflows/audit-selfcheck.yml b/.github/workflows/audit-selfcheck.yml
index 3954e6de..358585e9 100644
--- a/.github/workflows/audit-selfcheck.yml
+++ b/.github/workflows/audit-selfcheck.yml
@@ -21,7 +21,14 @@ jobs:
         run: pip install -e '.[dev,env]'
       - name: Run audit self-check
         run: labtrust audit-selfcheck --out ./audit_self_check
+      - name: Show audit diagnostics on failure
+        if: failure()
+        run: |
+          if test -f audit_self_check/AUDIT_SELF_CHECK.json; then
+            python -m json.tool audit_self_check/AUDIT_SELF_CHECK.json
+          fi
       - name: Upload audit self-check artifact
+        if: always()
         uses: actions/upload-artifact@v4
         with:
           name: audit_self_check
diff --git a/.github/workflows/pcs.yml b/.github/workflows/pcs.yml
index 5932357d..1c7e5c57 100644
--- a/.github/workflows/pcs.yml
+++ b/.github/workflows/pcs.yml
@@ -65,6 +65,8 @@ on:
 jobs:
   pcs:
     runs-on: ubuntu-latest
+    env:
+      PCS_CORE_PATH: ${{ github.workspace }}/pcs-core
     defaults:
       run:
         shell: bash
@@ -73,16 +75,47 @@ jobs:
       - uses: actions/checkout@v4
         with:
           repository: SentinelOps-CI/pcs-core
+          ref: 9c971f5f9da8a424924dd8f48d6a3b71a1009e1b
           path: pcs-core
       - uses: actions/checkout@v4
         with:
-          repository: SentinelOps-CI/pcs-bench
+          repository: fraware/pcs-bench
+          ref: 6092cccefee7841dfde1393e6881d433838a2252
           path: pcs-bench
       - name: Apply LabTrust pcs-core schema profiles
         run: python scripts/apply_pcs_core_labtrust_schema_profiles.py --pcs-core pcs-core
       - uses: actions/setup-python@v5
         with:
           python-version: "3.11"
+      - name: Prepare pinned pcs-core Python packaging metadata
+        run: |
+          python - <<'PY'
+          import shutil
+          import subprocess
+          from pathlib import Path
+
+          root = Path("pcs-core")
+          revision = subprocess.check_output(
+              ["git", "-C", str(root), "rev-parse", "HEAD"], text=True
+          ).strip()
+          expected = "9c971f5f9da8a424924dd8f48d6a3b71a1009e1b"
+          if revision != expected:
+              raise RuntimeError(f"Unexpected pcs-core revision: {revision}")
+          metadata = root / "python" / "pyproject.toml"
+          original = metadata.read_text(encoding="utf-8")
+          previous = 'readme = "../README.md"'
+          if original.count(previous) != 1:
+              raise RuntimeError("Unknown upstream packaging metadata; review needed")
+          source_readme = root / "README.md"
+          destination = root / "python" / "README.md"
+          if not source_readme.is_file() or destination.exists():
+              raise RuntimeError("Unexpected upstream README layout")
+          metadata.write_text(
+              original.replace(previous, 'readme = "README.md"'), encoding="utf-8"
+          )
+          shutil.copyfile(source_readme, destination)
+          print(f"Applied README-only packaging repair to pinned pcs-core {revision}")
+          PY
       - name: Install pcs-core, pcs-bench, and LabTrust-Gym
         run: |
           pip install -e "pcs-core/python"
diff --git a/.github/workflows/wheel-smoke.yml b/.github/workflows/wheel-smoke.yml
index fed7e4f0..ba25c3f9 100644
--- a/.github/workflows/wheel-smoke.yml
+++ b/.github/workflows/wheel-smoke.yml
@@ -26,9 +26,22 @@ jobs:
           python -m venv .venv-wheel
           .venv-wheel/bin/pip install dist/*.whl
           .venv-wheel/bin/pip install '.[env]'  # quick-eval uses env
-          .venv-wheel/bin/pip install pytest  # audit-selfcheck doctor (extras_dev) and risk_register_contract_gate
+          .venv-wheel/bin/pip install 'pytest>=8,<10' 'pytest-timeout>=2.3,<3'  # repo addopts includes --timeout
       - name: Audit selfcheck (doctor)
         run: .venv-wheel/bin/labtrust audit-selfcheck --out ./audit_out
+      - name: Show audit diagnostics on failure
+        if: failure()
+        run: |
+          if test -f audit_out/AUDIT_SELF_CHECK.json; then
+            python -m json.tool audit_out/AUDIT_SELF_CHECK.json
+          fi
+      - name: Upload audit diagnostics
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: wheel-audit-diagnostics
+          path: audit_out/AUDIT_SELF_CHECK.json
+          if-no-files-found: warn
       - name: Validate policy
         run: .venv-wheel/bin/labtrust validate-policy
       - name: Quick eval (smoke)
diff --git a/policy/schemas/risk_register_bundle.v0.1.schema.json b/policy/schemas/risk_register_bundle.v0.1.schema.json
index 456b17fe..a2059764 100644
--- a/policy/schemas/risk_register_bundle.v0.1.schema.json
+++ b/policy/schemas/risk_register_bundle.v0.1.schema.json
@@ -137,7 +137,7 @@
         "evidence_id": { "type": "string" },
         "type": {
           "type": "string",
-          "enum": ["security_suite", "coordination_study", "coordination_pack", "official_pack", "safety_case", "bundle_verification", "other"]
+          "enum": ["security_suite", "coordination_study", "coordination_pack", "coord_risk_run", "official_pack", "safety_case", "bundle_verification", "other"]
         },
         "path": { "type": "string", "description": "Repo-relative or run-relative path; empty or omitted for evidence gaps (status=missing)." },
         "label": { "type": "string" },
diff --git a/src/labtrust_gym/cli/main.py b/src/labtrust_gym/cli/main.py
index d1fa7f56..3abd111f 100644
--- a/src/labtrust_gym/cli/main.py
+++ b/src/labtrust_gym/cli/main.py
@@ -3081,6 +3081,8 @@ def _run_audit_selfcheck(args: argparse.Namespace) -> int:
 
     # 5. risk register contract gate (pytest)
     t0 = time.perf_counter()
+    contract_stdout = ""
+    contract_stderr = ""
     try:
         r = subprocess.run(
             [sys.executable, "-m", "pytest", "tests/test_risk_register_contract_gate.py", "-v", "--tb=short"],
@@ -3090,11 +3092,21 @@ def _run_audit_selfcheck(args: argparse.Namespace) -> int:
             timeout=120,
         )
         passed = r.returncode == 0
-    except Exception:
+        if not passed:
+            contract_stdout = r.stdout[-8000:]
+            contract_stderr = r.stderr[-4000:]
+    except Exception as exc:
         passed = False
-    steps.append(
-        {"name": "risk_register_contract_gate", "pass": passed, "duration_s": round(time.perf_counter() - t0, 3)}
-    )
+        contract_stderr = str(exc)
+    contract_step: dict[str, Any] = {
+        "name": "risk_register_contract_gate",
+        "pass": passed,
+        "duration_s": round(time.perf_counter() - t0, 3),
+    }
+    if not passed:
+        contract_step["output"] = contract_stdout
+        contract_step["error"] = contract_stderr
+    steps.append(contract_step)
     if not passed:
         all_pass = False
 
diff --git a/src/labtrust_gym/cli/pcs.py b/src/labtrust_gym/cli/pcs.py
index 33ab3311..976e3d57 100644
--- a/src/labtrust_gym/cli/pcs.py
+++ b/src/labtrust_gym/cli/pcs.py
@@ -10,25 +10,7 @@
 
 from labtrust_gym.cli.console import get_console
 from labtrust_gym.config import get_repo_root
-from labtrust_gym.pcs.attach_certificate import attach_certificate_files
 from labtrust_gym.pcs.demo import DEMO_SCENARIOS
-from labtrust_gym.pcs.demo import run_demo as execute_pcs_demo
-from labtrust_gym.pcs.export import export_pcs_bundle, export_runtime_receipt, export_trace
-from labtrust_gym.pcs.handoff import export_handoff_bundle
-from labtrust_gym.pcs.handoff_manifest import (
-    emit_handoff_manifest,
-    emit_handoff_to_certifyedge,
-    emit_handoff_to_pf,
-)
-from labtrust_gym.pcs.benchmark_cases import generate_benchmark_cases, verify_benchmark_cases
-from labtrust_gym.pcs.benchmark_reproducibility import benchmark_reproducibility
-from labtrust_gym.pcs.failure_gallery import generate_failure_gallery
-from labtrust_gym.pcs.regenerate_release_chain import regenerate_release_chain
-from labtrust_gym.pcs.regenerate_release_protocol import regenerate_release_protocol
-from labtrust_gym.pcs.release_protocol_producer import LABTRUST_PROTOCOL_PACKAGE_ARTIFACTS
-from labtrust_gym.pcs.release_fragment import emit_labtrust_release_fragment
-from labtrust_gym.pcs.status_policy import check_release_status_policy
-from labtrust_gym.pcs.verify_release_protocol import verify_release_protocol
 from labtrust_gym.pcs.validate import PcsValidationError, validate_artifact_file, validate_run_dir
 
 
@@ -46,6 +28,8 @@ def _print_validation_errors(errors: list[str]) -> None:
 
 
 def _run_demo(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.demo import run_demo as execute_pcs_demo
+
     name = args.demo
     out = _resolve_path(args.out) if args.out else None
     try:
@@ -73,6 +57,8 @@ def _run_demo(args: argparse.Namespace) -> int:
 
 
 def _run_export_trace(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.export import export_trace
+
     run_dir = _resolve_path(args.run)
     out = _resolve_path(args.out)
     if not run_dir.is_dir():
@@ -91,6 +77,8 @@ def _run_export_trace(args: argparse.Namespace) -> int:
 
 
 def _run_export_runtime_receipt(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.export import export_runtime_receipt
+
     run_dir = _resolve_path(args.run)
     out = _resolve_path(args.out)
     if not run_dir.is_dir():
@@ -109,6 +97,8 @@ def _run_export_runtime_receipt(args: argparse.Namespace) -> int:
 
 
 def _run_export_pcs(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.export import export_pcs_bundle
+
     run_dir = _resolve_path(args.run)
     out = _resolve_path(args.out)
     if not run_dir.is_dir():
@@ -127,6 +117,8 @@ def _run_export_pcs(args: argparse.Namespace) -> int:
 
 
 def _run_attach_certificate(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.attach_certificate import attach_certificate_files
+
     bundle_path = _resolve_path(args.bundle)
     cert_path = _resolve_path(args.certificate)
     out_path = _resolve_path(args.out)
@@ -167,6 +159,8 @@ def _run_validate_pcs(args: argparse.Namespace) -> int:
 
 
 def _run_emit_release_fragment(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.release_fragment import emit_labtrust_release_fragment
+
     release_dir = _resolve_path(args.release_dir)
     out = _resolve_path(args.out) if args.out else release_dir / "labtrust_release_fragment.json"
     try:
@@ -179,6 +173,8 @@ def _run_emit_release_fragment(args: argparse.Namespace) -> int:
 
 
 def _run_emit_handoff(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.handoff_manifest import emit_handoff_manifest
+
     out = _resolve_path(args.out)
     try:
         emit_handoff_manifest(
@@ -198,6 +194,8 @@ def _run_emit_handoff(args: argparse.Namespace) -> int:
 
 
 def _run_verify_release_protocol(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.verify_release_protocol import verify_release_protocol
+
     release_dir = _resolve_path(args.release_dir)
     pcs_core = _resolve_path(args.pcs_core) if args.pcs_core else None
     try:
@@ -216,6 +214,8 @@ def _run_verify_release_fixtures(args: argparse.Namespace) -> int:
 
 
 def _run_emit_handoff_to_certifyedge(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.handoff_manifest import emit_handoff_to_certifyedge
+
     out = _resolve_path(args.out)
     try:
         emit_handoff_to_certifyedge(
@@ -233,6 +233,8 @@ def _run_emit_handoff_to_certifyedge(args: argparse.Namespace) -> int:
 
 
 def _run_emit_handoff_to_pf(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.handoff_manifest import emit_handoff_to_pf
+
     out = _resolve_path(args.out)
     try:
         emit_handoff_to_pf(
@@ -248,6 +250,10 @@ def _run_emit_handoff_to_pf(args: argparse.Namespace) -> int:
 
 
 def _run_regenerate_release_protocol(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.regenerate_release_protocol import regenerate_release_protocol
+    from labtrust_gym.pcs.verify_release_protocol import verify_release_protocol
+    from labtrust_gym.pcs.release_protocol_producer import LABTRUST_PROTOCOL_PACKAGE_ARTIFACTS
+
     out = _resolve_path(args.out)
     pcs_core = _resolve_path(args.pcs_core) if args.pcs_core else None
     ce_spec = _resolve_path(args.certifyedge_spec) if args.certifyedge_spec else None
@@ -296,6 +302,8 @@ def _run_regenerate_release_protocol(args: argparse.Namespace) -> int:
 
 
 def _run_check_status_policy(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.status_policy import check_release_status_policy
+
     release_dir = _resolve_path(args.release_dir)
     try:
         profile_path = _resolve_path(args.workflow_profile) if getattr(args, "workflow_profile", None) else None
@@ -314,6 +322,8 @@ def _run_check_status_policy(args: argparse.Namespace) -> int:
 
 
 def _run_generate_benchmark_cases(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.benchmark_cases import generate_benchmark_cases
+
     out = _resolve_path(args.out)
     release_dir = _resolve_path(args.release_dir) if args.release_dir else None
     validate_root = (
@@ -361,6 +371,8 @@ def _run_generate_benchmark_cases(args: argparse.Namespace) -> int:
 
 
 def _run_benchmark_reproducibility(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.benchmark_reproducibility import benchmark_reproducibility
+
     out = _resolve_path(args.out)
     pcs_core = _resolve_path(args.pcs_core) if args.pcs_core else None
     release_dir = _resolve_path(args.release_dir) if args.release_dir else None
@@ -442,6 +454,8 @@ def _resolve_pcs_core_root(explicit: Path | None) -> Path | None:
 
 
 def _run_verify_benchmark_cases(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.benchmark_cases import verify_benchmark_cases
+
     root = _resolve_path(args.benchmark_dir)
     pcs_core = _resolve_pcs_core_root(
         _resolve_path(args.pcs_core) if getattr(args, "pcs_core", None) else None
@@ -495,6 +509,8 @@ def _run_verify_benchmark_cases(args: argparse.Namespace) -> int:
 
 
 def _run_generate_failure_gallery(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.failure_gallery import generate_failure_gallery
+
     out = _resolve_path(args.out)
     release_dir = _resolve_path(args.release_dir) if args.release_dir else None
     try:
@@ -518,6 +534,8 @@ def _run_generate_failure_gallery(args: argparse.Namespace) -> int:
 
 
 def _run_regenerate_release_chain(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.regenerate_release_chain import regenerate_release_chain
+
     out = _resolve_path(args.out)
     pcs_core = _resolve_path(args.pcs_core) if args.pcs_core else None
     ce_spec = _resolve_path(args.certifyedge_spec) if args.certifyedge_spec else None
@@ -538,6 +556,8 @@ def _run_regenerate_release_chain(args: argparse.Namespace) -> int:
 
 
 def _run_export_handoff(args: argparse.Namespace) -> int:
+    from labtrust_gym.pcs.handoff import export_handoff_bundle
+
     out = _resolve_path(args.out)
     try:
         manifest = export_handoff_bundle(
diff --git a/src/labtrust_gym/pcs/demo.py b/src/labtrust_gym/pcs/demo.py
index 1e973a0b..00e1fc3a 100644
--- a/src/labtrust_gym/pcs/demo.py
+++ b/src/labtrust_gym/pcs/demo.py
@@ -5,8 +5,6 @@
 from pathlib import Path
 
 from labtrust_gym.config import get_repo_root
-from labtrust_gym.pcs.workflows.qc_release import FAILURE_CASES, run_failure_demo
-from labtrust_gym.pcs.workflows.registry import get_workflow
 
 DEMO_SCENARIOS: dict[str, tuple[str, str]] = {
     "qc-release": ("valid_workflow.yaml", "runs/qc-release"),
@@ -33,6 +31,10 @@ def run_demo(
     policy_root: Path | None = None,
     deterministic: bool | None = None,
 ) -> Path:
+    # Import workflow engines only when a PCS demo is actually requested.
+    from labtrust_gym.pcs.workflows.qc_release import FAILURE_CASES, run_failure_demo
+    from labtrust_gym.pcs.workflows.registry import get_workflow
+
     if demo_name not in DEMO_SCENARIOS:
         raise ValueError(f"unknown demo {demo_name!r}; choose from {list(DEMO_SCENARIOS)}")
     root = policy_root or get_repo_root()
diff --git a/src/labtrust_gym/verifier_assurance/reward/composition.py b/src/labtrust_gym/verifier_assurance/reward/composition.py
index 98058d5b..b1d119ee 100644
--- a/src/labtrust_gym/verifier_assurance/reward/composition.py
+++ b/src/labtrust_gym/verifier_assurance/reward/composition.py
@@ -6,6 +6,7 @@
 import hashlib
 from typing import Any, Mapping
 
+from labtrust_gym.config import get_repo_root
 from labtrust_gym.errors import PolicyLoadError
 from labtrust_gym.policy.loader import load_json, validate_against_schema
 from labtrust_gym.util.json_utils import canonical_json
@@ -26,20 +27,9 @@
     "safety_violation_penalty",
 )
 
-_SCHEMA_PATH = (
-    Path(__file__).resolve().parents[4]
-    / "policy"
-    / "schemas"
-    / "verifier_assurance"
-    / "RewardCompositionPolicy.v1.schema.json"
-)
-_ENVELOPE_SCHEMA_PATH = (
-    Path(__file__).resolve().parents[4]
-    / "policy"
-    / "schemas"
-    / "pcs"
-    / "RewardEvidenceEnvelope.v1.schema.json"
-)
+def _reward_schema_path(*parts: str) -> Path:
+    """Resolve a policy schema using package, checkout, or explicit configuration."""
+    return get_repo_root() / "policy" / "schemas" / Path(*parts)
 
 
 class RewardCompositionError(ValueError):
@@ -52,8 +42,9 @@ def compute_policy_digest(policy: dict[str, Any]) -> str:
 
 
 def validate_composition_policy(policy: dict[str, Any]) -> dict[str, Any]:
+    schema_path = _reward_schema_path("verifier_assurance", "RewardCompositionPolicy.v1.schema.json")
     try:
-        validate_against_schema(policy, load_json(_SCHEMA_PATH), path=_SCHEMA_PATH)
+        validate_against_schema(policy, load_json(schema_path), path=schema_path)
     except PolicyLoadError as exc:
         raise RewardCompositionError(str(exc)) from exc
     weights = policy.get("weights") or {}
@@ -209,8 +200,9 @@ def build_reward_evidence_envelope(
         "public_decision": public_decision,
         "claim_boundary": CLAIM_BOUNDARY,
     }
+    schema_path = _reward_schema_path("pcs", "RewardEvidenceEnvelope.v1.schema.json")
     try:
-        validate_against_schema(envelope, load_json(_ENVELOPE_SCHEMA_PATH), path=_ENVELOPE_SCHEMA_PATH)
+        validate_against_schema(envelope, load_json(schema_path), path=schema_path)
     except PolicyLoadError as exc:
         raise RewardCompositionError(str(exc)) from exc
     return envelope
diff --git a/tests/fixtures/risk_register_bundle_ui_fixtures.v0.1.json b/tests/fixtures/risk_register_bundle_ui_fixtures.v0.1.json
index 14f64f19..0208a52b 100644
--- a/tests/fixtures/risk_register_bundle_ui_fixtures.v0.1.json
+++ b/tests/fixtures/risk_register_bundle_ui_fixtures.v0.1.json
@@ -619,7 +619,7 @@
     "method_risk_matrix": "8a44089fce4ac7e95cdc1040987fc5bddfd7d4b4fc6348dbc7fc851a753fd195",
     "risk_registry": "74c6a81361dd50e0d291e8d21cdb21ce0d2996bbf5f56ae93501d8dc51280b45",
     "safety_case_claims": "8de856fa7fb6b2b3a1aed725971cb2fc9cfd20b2210dc9a1565e6445a998d749",
-    "security_attack_suite": "b994010c6d13810b66621a8c3169d13aed8cc144584af50ef66976d7bc18c1e0"
+    "security_attack_suite": "bada4e173a03dff42b598203cf09df2705699a16a7c81cd73b700726e5fdc829"
   },
   "reproduce": [
     {
@@ -1283,4 +1283,4 @@
       "typical_failure_mode": "Malicious input to tool; abuse of optional parameters; bypass of checks."
     }
   ]
-}
\ No newline at end of file
+}
diff --git a/tests/test_cli_optional_pcs_import_boundary.py b/tests/test_cli_optional_pcs_import_boundary.py
new file mode 100644
index 00000000..12f21a2f
--- /dev/null
+++ b/tests/test_cli_optional_pcs_import_boundary.py
@@ -0,0 +1,27 @@
+"""Regression: core CLI initialization must not import the optional PCS package."""
+
+from __future__ import annotations
+
+import subprocess
+import sys
+
+
+def test_cli_help_works_without_pcs_core() -> None:
+    """The default CLI command registry is available with only base dependencies."""
+    script = """
+import sys
+# Block importing pcs_core even if a developer installed the PCS optional extra.
+sys.modules["pcs_core"] = None
+from labtrust_gym.cli.main import main
+sys.argv = ["labtrust", "--help"]
+main()
+"""
+    result = subprocess.run(
+        [sys.executable, "-c", script],
+        text=True,
+        capture_output=True,
+        timeout=40,
+        check=False,
+    )
+    assert result.returncode == 0, result.stderr
+    assert "usage:" in result.stdout.lower()
diff --git a/tests/test_no_placeholders.py b/tests/test_no_placeholders.py
index 77c0a9be..cd0c64a1 100644
--- a/tests/test_no_placeholders.py
+++ b/tests/test_no_placeholders.py
@@ -162,3 +162,56 @@ def test_clean_tree_no_violations(tmp_path: Path) -> None:
     )
     violations = scan_root(tmp_path)
     assert not violations
+
+
+def test_exception_types_are_distinct_from_unimplemented_functions(tmp_path: Path) -> None:
+    """A handler and a custom exception class are legal; an unfinished function fails."""
+    _write_tree(
+        tmp_path,
+        {
+            "src/types.py": (
+                "class Unavailable(NotImplementedError):\n    pass\n"
+                "def handle():\n"
+                "    try:\n        return 1\n"
+                "    except (ValueError, NotImplementedError):\n        return 0\n"
+            ),
+            "src/unimplemented.py": "def unfinished():\n    raise NotImplementedError('TODO')\n",
+        },
+    )
+    problems = [
+        _norm_path(p, tmp_path)
+        for p, _, token in scan_root(tmp_path)
+        if token == "NotImplementedError"
+    ]
+    assert problems == ["src/unimplemented.py"]
+
+
+def test_template_scaffold_is_excluded_without_excluding_production(tmp_path: Path) -> None:
+    _write_tree(
+        tmp_path,
+        {
+            "templates/pcs_workflow_template/workflow.py": "raise NotImplementedError('template')\n",
+            "src/actual.py": "raise NotImplementedError('production')\n",
+        },
+    )
+    problems = [_norm_path(p, tmp_path) for p, _, _ in scan_root(tmp_path)]
+    assert problems == ["src/actual.py"]
+
+
+def test_schema_fields_are_precisely_exempt_and_other_policy_markers_fail(
+    tmp_path: Path,
+) -> None:
+    _write_tree(
+        tmp_path,
+        {
+            "policy/schemas/pcs/LabtrustBenchmarkExtension.v0.schema.json": (
+                '"placeholder_fields": {\n'
+                '"description": "Documents placeholder benchmark_case fields when pcs-core expects null on valid cases.",\n'
+                '"description": "Unexpected placeholder claim.",\n'
+            ),
+        },
+    )
+    failures = [
+        (line, token) for _, line, token in scan_root(tmp_path) if token == "placeholder"
+    ]
+    assert failures == [(3, "placeholder")]
diff --git a/tests/test_risk_register_bundle.py b/tests/test_risk_register_bundle.py
index 82e523bc..f5a5112e 100644
--- a/tests/test_risk_register_bundle.py
+++ b/tests/test_risk_register_bundle.py
@@ -271,3 +271,26 @@ def test_coord_summary_includes_metrics(tmp_path: Path) -> None:
     assert len(rows) >= 1
     row = rows[0]
     assert "sec.attack_success_rate" in row or "robustness.resilience_score" in row
+
+def test_coord_risk_evidence_type_is_supported_without_accepting_unknown_types() -> None:
+    """Preserve the evidence type emitted by the strict coordination-risk producer."""
+    root = _repo_root()
+    bundle = build_risk_register_bundle(
+        root,
+        run_dirs=[],
+        include_generated_at=False,
+        include_git_hash=False,
+    )
+    entry = {
+        "evidence_id": "EVID-COORD-RISK-SYNTHETIC",
+        "type": "coord_risk_run",
+        "status": "present",
+        "path": "runs/coord_risk_fixture/coord_risk_results.json",
+        "risk_ids": ["R-SYS-001"],
+    }
+    bundle["evidence"].append(entry)
+    assert validate_bundle_against_schema(bundle, root) == []
+
+    entry["type"] = "unrecognized_evidence_class"
+    assert validate_bundle_against_schema(bundle, root)
+
diff --git a/tests/verifier_assurance/test_va03_reward_composition.py b/tests/verifier_assurance/test_va03_reward_composition.py
index f84fb0e2..3641150f 100644
--- a/tests/verifier_assurance/test_va03_reward_composition.py
+++ b/tests/verifier_assurance/test_va03_reward_composition.py
@@ -3,11 +3,13 @@
 from __future__ import annotations
 
 import copy
+from pathlib import Path
 
 import pytest
 
 from labtrust_gym.verifier_assurance.reward.composition import (
     COMPONENT_VOCAB,
+    _reward_schema_path,
     RewardComposer,
     RewardCompositionError,
     apply_legacy_rewards,
@@ -116,3 +118,17 @@ def test_emit_component_breakdown_when_enabled() -> None:
     assert rewards["ops_0"] == 1.0
     assert breakdown is not None
     assert breakdown["ops_0"]["operational_success"] == 1.0
+
+def test_reward_schema_paths_respect_explicit_policy_root(monkeypatch: pytest.MonkeyPatch) -> None:
+    """An installed module must resolve schemas through its configured policy root."""
+    repo = Path(__file__).resolve().parents[2]
+    monkeypatch.setenv("LABTRUST_POLICY_DIR", str(repo / "policy"))
+    composition_schema = _reward_schema_path(
+        "verifier_assurance", "RewardCompositionPolicy.v1.schema.json"
+    )
+    envelope_schema = _reward_schema_path("pcs", "RewardEvidenceEnvelope.v1.schema.json")
+    assert composition_schema.is_file()
+    assert envelope_schema.is_file()
+    assert composition_schema.is_relative_to(repo / "policy")
+    assert envelope_schema.is_relative_to(repo / "policy")
+
diff --git a/tools/no_placeholders.py b/tools/no_placeholders.py
index 73fe2622..8c642e55 100644
--- a/tools/no_placeholders.py
+++ b/tools/no_placeholders.py
@@ -8,6 +8,7 @@
 
 from __future__ import annotations
 
+import ast
 import re
 import subprocess
 import sys
@@ -85,7 +86,12 @@ def _is_excluded_path(rel_path: str) -> bool:
 SECRET_SCRUBBER_NAME = "secret_scrubber.py"
 
 # Files that define or test the gate itself: skip to avoid self-fail
-EXCLUDE_FILES = {"tools/no_placeholders.py", "tests/test_no_placeholders.py"}
+EXCLUDE_FILES = {
+    "tools/no_placeholders.py",
+    "tests/test_no_placeholders.py",
+    # This published integration scaffold intentionally contains abstract method bodies.
+    "templates/pcs_workflow_template/workflow.py",
+}
 
 # (rel_path, pattern): allow these global-fail patterns in config (e.g. coverage exclude_lines)
 # agent_api.py: base class default act() raises NotImplementedError; subclasses must implement.
@@ -135,15 +141,53 @@ def _is_under_tests_for_skip(rel_path: str) -> bool:
     return rel_path.startswith("tests/")
 
 
+def _non_stub_not_implemented_lines(path: Path, lines: list[str]) -> set[int]:
+    """Recognize exception *types* in handlers and class bases, never stub raises."""
+    if path.suffix.lower() != ".py":
+        return set()
+    try:
+        tree = ast.parse("\n".join(lines) + "\n")
+    except SyntaxError:
+        return set()
+
+    def exception_names(node: ast.expr | None) -> list[ast.Name]:
+        if isinstance(node, ast.Name):
+            return [node]
+        if isinstance(node, ast.Tuple):
+            return [name for item in node.elts for name in exception_names(item)]
+        return []
+
+    allowed: set[int] = set()
+    for node in ast.walk(tree):
+        if isinstance(node, ast.ExceptHandler):
+            refs = exception_names(node.type)
+        elif isinstance(node, ast.ClassDef):
+            refs = [name for base in node.bases for name in exception_names(base)]
+        else:
+            continue
+        allowed.update(
+            name.lineno for name in refs if name.id == "NotImplementedError"
+        )
+    return allowed
+
+
 def _check_global_fails(path: Path, root: Path, lines: list[str]) -> list[tuple[int, str]]:
     violations: list[tuple[int, str]] = []
     rel_path = _norm_path(path, root)
     if _is_under_tests_for_skip(rel_path):
         return []
+    allowed_exception_lines = _non_stub_not_implemented_lines(path, lines)
     for i, line in enumerate(lines, start=1):
         for pat in GLOBAL_FAIL_PATTERNS:
-            if pat in line and (rel_path, pat) not in ALLOWED_GLOBAL_VIOLATIONS:
-                violations.append((i, pat))
+            if pat not in line or (rel_path, pat) in ALLOWED_GLOBAL_VIOLATIONS:
+                continue
+            if (
+                pat == "NotImplementedError"
+                and i in allowed_exception_lines
+                and not re.search(r"\braise\s+NotImplementedError\b", line)
+            ):
+                continue
+            violations.append((i, pat))
     return violations
 
 
@@ -174,7 +218,15 @@ def _check_word_placeholder_stub(path: Path, root: Path, lines: list[str]) -> li
         if _is_secret_scrubber(rel_path) and "placeholder" in line.lower():
             continue
         if "placeholder" in line.lower() and not allow_placeholder:
-            violations.append((i, "placeholder"))
+            # Exact schema declarations describe compatibility metadata;
+            # unrelated unfinished policy text still fails the scan.
+            canonical_schema = "policy/schemas/pcs/LabtrustBenchmarkExtension.v0.schema.json"
+            recognized_fields = {
+                '"placeholder_fields": {',
+                '"description": "Documents placeholder benchmark_case fields when pcs-core expects null on valid cases.",',
+            }
+            if rel_path != canonical_schema or line.strip() not in recognized_fields:
+                violations.append((i, "placeholder"))
         # In tests/, allow capital-S "Stub" (e.g. StubTask); fail only on lowercase "stub"
         if "stub" in line.lower() and not allow_stub:
             if rel_path.startswith("tests/") and re.search(r"\bstub\b", line):