diff --git a/zenith/src/zenith_harness/attention.py b/zenith/src/zenith_harness/attention.py
index 7399d8d..da5e407 100644
--- a/zenith/src/zenith_harness/attention.py
+++ b/zenith/src/zenith_harness/attention.py
@@ -114,11 +114,23 @@ def _gate_report(
     failed_items = failed_items or []
     validator_verdicts = validator_verdicts or {}
     missing_items = missing_items or {}
-    lines = [
-        f"Gate report from {gate.id}",
-        f"cleared: {cleared}",
-        f"targets: {', '.join(gate.targets) if gate.targets else '(none)'}",
-    ]
+    title = (
+        f"Gate checkpoint from {gate.id}"
+        if cleared
+        else f"Gate report from {gate.id}"
+    )
+    lines = [title]
+    if cleared:
+        lines.append(
+            "checkpoint: passing gate already cleared; continue acknowledges "
+            "the checkpoint and does not clear a failed task or gate"
+        )
+    lines.extend(
+        [
+            f"cleared: {cleared}",
+            f"targets: {', '.join(gate.targets) if gate.targets else '(none)'}",
+        ]
+    )
     if reason:
         lines.append(f"reason: {reason}")
     if failed_items:
diff --git a/zenith/src/zenith_harness/bundled/prompts/orchestrator/system_prompt.md b/zenith/src/zenith_harness/bundled/prompts/orchestrator/system_prompt.md
index 22f6589..45e4689 100644
--- a/zenith/src/zenith_harness/bundled/prompts/orchestrator/system_prompt.md
+++ b/zenith/src/zenith_harness/bundled/prompts/orchestrator/system_prompt.md
@@ -308,7 +308,7 @@ Use decision actions narrowly:
 
 Patch the earliest invalid artifact. Do not create another fix task when the real defect is vague scope, missing inventory coverage, broad contract, stale `mission.md`, wrong task grouping, obsolete skill, weak validator method, missing setup, or bad oracle. If pass/fail criteria changed, old evidence may no longer prove the assertion; plan revalidation.
 
-Be especially careful with `continue`: for failed task or failed gate reports, runtime may treat `continue` as clearing that task. Do not use it to skip failed implementation, missing validator evidence, unresolved gate dissent, merge conflicts, or a terminal-review gap.
+Be especially careful with `continue`: for failed task or failed gate reports, runtime may treat `continue` as clearing that task. Passing gate checkpoints are different: a report headed `Gate checkpoint` already has `cleared: True`, and `continue` only acknowledges the checkpoint. Do not use `continue` to skip failed implementation, missing validator evidence, unresolved gate dissent, merge conflicts, or a terminal-review gap.
 
 When adding new assertion ids in a patch, write matching `contract/<ID>.md` files before calling `decide_attention`. When changing a task body, skill, targets, or dependencies, supersede the task rather than pretending it was edited in place. Do not silently mutate cleared or running work.
 
diff --git a/zenith/src/zenith_harness/controller.py b/zenith/src/zenith_harness/controller.py
index 65f5468..c58d82c 100644
--- a/zenith/src/zenith_harness/controller.py
+++ b/zenith/src/zenith_harness/controller.py
@@ -11,7 +11,9 @@
 from __future__ import annotations
 
 from dataclasses import dataclass
+from pathlib import Path
 
+from .assets import AssetLoader
 from .config import HarnessConfig
 from .coordinator import MissionCoordinator
 from .dispatcher import NodeDispatcher, TerminalReviewer
@@ -27,7 +29,6 @@
     MissionPlanning,
     MissionRunning,
     ProjectState,
-    Task,
     TaskList,
     TaskListPatch,
     TaskStateFile,
@@ -36,6 +37,7 @@
 from .task_list_patch import apply_patch
 from .task_validation import (
     ValidationError,
+    check_skill_names,
     parse_contract_dir,
     validate_task_list_submission,
 )
@@ -101,6 +103,8 @@ def submit_plan(self, project_id: str, task_list: TaskList) -> Envelope:
                 details=parse_errs,
             )
         errs = validate_task_list_submission(ids, task_list)
+        if not errs:
+            errs = self._validate_task_skills(project_id, task_list)
         if errs:
             raise ToolError(
                 "invalid_task_list", "task list validation failed", details=errs
@@ -357,6 +361,11 @@ def _apply_task_list_patch(
             raise ToolError(
                 "invalid_patch", "patch validation failed", details=errs
             )
+        skill_errs = self._validate_task_skills(project_id, patched_tl)
+        if skill_errs:
+            raise ToolError(
+                "invalid_patch", "patch validation failed", details=skill_errs
+            )
         self.store.save_task_list(project_id, mid, patched_tl)
         self.store.save_task_state(project_id, mid, patched_state)
         cs = self.store.load_contract_state(project_id, mid)
@@ -406,6 +415,21 @@ def _require_state(self, project_id: str) -> ProjectState:
             raise ToolError("not_found", f"project {project_id!r} has no state")
         return state
 
+    def _validate_task_skills(
+        self,
+        project_id: str,
+        task_list: TaskList,
+    ) -> list[ValidationError]:
+        return check_skill_names(task_list, self._available_skill_names(project_id))
+
+    def _available_skill_names(self, project_id: str) -> set[str]:
+        names: set[str] = set()
+        for skill in AssetLoader(self.config).list_skills(project_id):
+            if skill.name:
+                names.add(skill.name)
+            names.add(Path(skill.path).parent.name)
+        return names
+
 
 def _make_pending():
     from .models import ContractStateEntry
diff --git a/zenith/src/zenith_harness/task_validation.py b/zenith/src/zenith_harness/task_validation.py
index d63f341..7f5bb12 100644
--- a/zenith/src/zenith_harness/task_validation.py
+++ b/zenith/src/zenith_harness/task_validation.py
@@ -118,6 +118,31 @@ def check_task_shape(tl: TaskList) -> list[ValidationError]:
     return errors
 
 
+def check_skill_names(
+    tl: TaskList,
+    available_skills: set[str] | list[str] | tuple[str, ...],
+) -> list[ValidationError]:
+    """Non-gate task skills must resolve to an available skill name."""
+    available = set(available_skills)
+    errors: list[ValidationError] = []
+    for task in tl.tasks:
+        if task.type == "gate" or not task.skill:
+            continue
+        if task.skill not in available:
+            if available:
+                detail = (
+                    f"{task.id} references unknown skill {task.skill!r}; "
+                    f"available: {', '.join(sorted(available))}"
+                )
+            else:
+                detail = (
+                    f"{task.id} references unknown skill {task.skill!r}; "
+                    "no skills are available"
+                )
+            errors.append(ValidationError("unknown_skill", detail))
+    return errors
+
+
 def check_deps_resolve(tl: TaskList) -> list[ValidationError]:
     """All `depends_on` ids reference declared tasks; no self-loop."""
     errors: list[ValidationError] = []
@@ -295,6 +320,7 @@ def upstream_tasks(
     "parse_contract_dir",
     "check_task_ids",
     "check_task_shape",
+    "check_skill_names",
     "check_deps_resolve",
     "check_acyclic",
     "check_coverage",
diff --git a/zenith/tests/test_cli.py b/zenith/tests/test_cli.py
index 9656f54..17de192 100644
--- a/zenith/tests/test_cli.py
+++ b/zenith/tests/test_cli.py
@@ -94,7 +94,7 @@ def test_codex_writes_codex_config(
         assert server["args"] == _expected_mcp_server_args()
         assert f"Initialized v5 project workspace at {workspace}" in r.output
         assert "Start your agent from the initialized project workspace" in r.output
-        assert "Read .codex/orchestrator_prompt.md and use Zenith to run this mission." in r.output
+        assert ".codex/orchestrator_prompt.md and treat it as your primary role" in r.output
 
     def test_claude_init_writes_runtime_validator_env_names(
         self, runner: CliRunner, workspace: Path, env: dict[str, str]
diff --git a/zenith/tests/test_coordinator.py b/zenith/tests/test_coordinator.py
index 71e8697..5419719 100644
--- a/zenith/tests/test_coordinator.py
+++ b/zenith/tests/test_coordinator.py
@@ -56,7 +56,11 @@ def _task(
     body: str = "body",
 ) -> Task:
     if skill is None and ttype != "gate":
-        skill = "s"
+        skill = (
+            "scrutiny-validator"
+            if ttype == "validate"
+            else "engineering-mission-playbook"
+        )
     return Task(
         id=tid,
         type=ttype,  # type: ignore[arg-type]
@@ -71,7 +75,7 @@ def _simple_tl() -> TaskList:
     return TaskList(
         tasks=[
             _task("w1", "work", ["VAL-001"]),
-            _task("v1", "validate", ["VAL-001"], skill="aud", depends_on=["w1"]),
+            _task("v1", "validate", ["VAL-001"], depends_on=["w1"]),
             _task("g1", "gate", ["VAL-001"], depends_on=["v1"]),
         ]
     )
@@ -81,7 +85,7 @@ def _validate_no_gate_tl() -> TaskList:
     return TaskList(
         tasks=[
             _task("w1", "work", ["VAL-001"]),
-            _task("v1", "validate", ["VAL-001"], skill="aud", depends_on=["w1"]),
+            _task("v1", "validate", ["VAL-001"], depends_on=["w1"]),
         ]
     )
 
@@ -129,11 +133,21 @@ def responder(req: DispatchRequest) -> NodeHandoff:
         assert env.state.state == "attention_needed"
         items = controller.store.load_attention(pid)
         assert items and items[0].kind == "gate_checkpoint"
+        assert items[0].report.startswith("Gate checkpoint from g1")
+        assert "passing gate already cleared" in items[0].report
+        assert (
+            controller.store.load_task_state(pid, "mission-001").status_of("g1")
+            == "cleared"
+        )
 
         env = controller.decide_attention(
             pid, [Decision(item_id=items[0].id, action="continue")]
         )
         assert env.state.state in ("mission_running", "done")
+        assert (
+            controller.store.load_task_state(pid, "mission-001").status_of("g1")
+            == "cleared"
+        )
 
         env = controller.advance_project(pid)
         assert env.state.state == "mission_running"
@@ -182,6 +196,69 @@ def responder(req: DispatchRequest) -> NodeHandoff:
         assert codex_skills.is_dir()
         assert not codex_skills.is_symlink()
 
+    def test_submit_plan_rejects_unknown_skill(
+        self, config: HarnessConfig, workspace: Path
+    ) -> None:
+        controller = ProjectController(
+            config,
+            MockDispatcher(lambda r: WorkHandoff(node_id=r.task.id, done=True, report="")),
+            MockTerminalReviewer(TerminalReviewHandoff(done=True, report="")),
+        )
+        pid = _seed_project(controller, workspace)
+        plan = TaskList(
+            tasks=[
+                _task(
+                    "w1",
+                    "work",
+                    ["VAL-001"],
+                    skill="missing-worker-skill",
+                ),
+                _task("v1", "validate", ["VAL-001"], depends_on=["w1"]),
+                _task("g1", "gate", ["VAL-001"], depends_on=["v1"]),
+            ]
+        )
+
+        with pytest.raises(ToolError) as err:
+            controller.submit_plan(pid, plan)
+
+        assert err.value.code == "invalid_task_list"
+        assert err.value.details is not None
+        assert any(detail.code == "unknown_skill" for detail in err.value.details)
+
+    def test_patch_rejects_unknown_skill(
+        self, config: HarnessConfig, workspace: Path
+    ) -> None:
+        controller = ProjectController(
+            config,
+            MockDispatcher(lambda r: WorkHandoff(node_id=r.task.id, done=True, report="")),
+            MockTerminalReviewer(TerminalReviewHandoff(done=True, report="")),
+        )
+        pid = _seed_project(controller, workspace)
+        controller.submit_plan(pid, _simple_tl())
+        contract_dir = controller.store.ensure_contract_dir(pid, "mission-001")
+        (contract_dir / "NEW-001.md").write_text("# NEW-001\n\nStatement body.\n")
+        patch = TaskListPatch(
+            add_items=["NEW-001"],
+            add=[
+                _task(
+                    "w2",
+                    "work",
+                    ["NEW-001"],
+                    skill="missing-worker-skill",
+                    depends_on=["g1"],
+                ),
+                _task("v2", "validate", ["NEW-001"], depends_on=["w2"]),
+                _task("g2", "gate", ["NEW-001"], depends_on=["v2"]),
+            ],
+        )
+
+        with pytest.raises(ToolError) as err:
+            controller._apply_task_list_patch(pid, "mission-001", patch)
+
+        assert err.value.code == "invalid_patch"
+        assert err.value.details is not None
+        assert any(detail.code == "unknown_skill" for detail in err.value.details)
+
 
 class TestNodeFailedFlow:
     def test_work_failure_raises_attention(
@@ -285,8 +362,8 @@ def _two_validator_tl() -> TaskList:
         return TaskList(
             tasks=[
                 _task("w1", "work", ["VAL-001"]),
-                _task("v-scrutiny", "validate", ["VAL-001"], skill="aud", depends_on=["w1"]),
-                _task("v-user-surface", "validate", ["VAL-001"], skill="aud", depends_on=["w1"]),
+                _task("v-scrutiny", "validate", ["VAL-001"], skill="scrutiny-validator", depends_on=["w1"]),
+                _task("v-user-surface", "validate", ["VAL-001"], skill="user-testing-validator", depends_on=["w1"]),
                 _task("g1", "gate", ["VAL-001"], depends_on=["v-scrutiny", "v-user-surface"]),
             ]
         )
@@ -537,7 +614,7 @@ def responder(req: DispatchRequest) -> NodeHandoff:
             add_items=["NEW-001"],
             add=[
                 _task("w2", "work", ["NEW-001"], depends_on=["g1"]),
-                _task("v2", "validate", ["NEW-001"], skill="aud", depends_on=["w2"]),
+                _task("v2", "validate", ["NEW-001"], depends_on=["w2"]),
                 _task("g2", "gate", ["NEW-001"], depends_on=["v2"]),
             ],
         )
@@ -589,7 +666,7 @@ def responder(req: DispatchRequest) -> NodeHandoff:
 
         submitted = controller.submit_plan(pid, _simple_tl())
         assert submitted.dag is not None
-        assert "    w1  [work:s]  pending  → VAL-001  ← (root)" in submitted.dag
+        assert "    w1  [work:engineering-mission-playbook]  pending  → VAL-001  ← (root)" in submitted.dag
         assert "focus-subgraph" not in submitted.dag
 
         inspected = controller.inspect_project(pid)
diff --git a/zenith/tests/test_coordinator_parallel.py b/zenith/tests/test_coordinator_parallel.py
index a290e22..6c78378 100644
--- a/zenith/tests/test_coordinator_parallel.py
+++ b/zenith/tests/test_coordinator_parallel.py
@@ -40,7 +40,13 @@ def config(harness_home: Path) -> HarnessConfig:
 
 
 def _task(tid: str, target: str) -> Task:
-    return Task(id=tid, type="work", body="b", targets=[target], skill="s")
+    return Task(
+        id=tid,
+        type="work",
+        body="b",
+        targets=[target],
+        skill="engineering-mission-playbook",
+    )
 
 
 def _write_contract(store: ProjectStore, pid: str, mission_id: str, assertion: str) -> None:
@@ -114,7 +120,7 @@ def responder(req: DispatchRequest) -> WorkHandoff:
                     type="work",
                     body="b",
                     targets=["EXP-A"],
-                    skill="s",
+                    skill="engineering-mission-playbook",
                     auto_merge=False,
                 )
             ]
@@ -175,7 +181,7 @@ def responder(req: DispatchRequest) -> WorkHandoff | ValidateHandoff:
                     type="work",
                     body="later work",
                     targets=[],
-                    skill="s",
+                    skill="engineering-mission-playbook",
                     depends_on=["w1"],
                 ),
                 Task(
@@ -183,7 +189,7 @@ def responder(req: DispatchRequest) -> WorkHandoff | ValidateHandoff:
                     type="validate",
                     body="audit implementation",
                     targets=["VAL-A"],
-                    skill="aud",
+                    skill="scrutiny-validator",
                     depends_on=["w1"],
                 ),
                 Task(
@@ -191,7 +197,7 @@ def responder(req: DispatchRequest) -> WorkHandoff | ValidateHandoff:
                     type="validate",
                     body="audit user surface",
                     targets=["VAL-A"],
-                    skill="aud",
+                    skill="user-testing-validator",
                     depends_on=["w1"],
                 ),
                 Task(
@@ -259,7 +265,7 @@ def responder(req: DispatchRequest) -> WorkHandoff | ValidateHandoff:
                     type="work",
                     body="later work",
                     targets=[],
-                    skill="s",
+                    skill="engineering-mission-playbook",
                     depends_on=["w1"],
                 ),
                 Task(
@@ -267,7 +273,7 @@ def responder(req: DispatchRequest) -> WorkHandoff | ValidateHandoff:
                     type="validate",
                     body="already audited",
                     targets=["VAL-A"],
-                    skill="aud",
+                    skill="scrutiny-validator",
                     depends_on=["w1"],
                 ),
                 Task(
@@ -275,7 +281,7 @@ def responder(req: DispatchRequest) -> WorkHandoff | ValidateHandoff:
                     type="validate",
                     body="remaining audit",
                     targets=["VAL-A"],
-                    skill="aud",
+                    skill="user-testing-validator",
                     depends_on=["w1"],
                 ),
                 Task(
diff --git a/zenith/tests/test_runnable_selection.py b/zenith/tests/test_runnable_selection.py
index a52e8fa..bd9ec91 100644
--- a/zenith/tests/test_runnable_selection.py
+++ b/zenith/tests/test_runnable_selection.py
@@ -36,7 +36,11 @@ def _task(
     depends_on: list[str] | None = None,
 ) -> Task:
     if skill is None and ttype != "gate":
-        skill = "s"
+        skill = (
+            "scrutiny-validator"
+            if ttype == "validate"
+            else "engineering-mission-playbook"
+        )
     return Task(
         id=tid,
         type=ttype,  # type: ignore[arg-type]
@@ -191,8 +195,8 @@ def responder(req: DispatchRequest) -> NodeHandoff:
         controller.submit_plan(pid, TaskList(tasks=[
             _task("a", "work", ["VAL-A"]),
             _task("b", "work", ["VAL-B"]),
-            _task("va", "validate", ["VAL-A"], skill="aud", depends_on=["a"]),
-            _task("vb", "validate", ["VAL-B"], skill="aud", depends_on=["b"]),
+            _task("va", "validate", ["VAL-A"], depends_on=["a"]),
+            _task("vb", "validate", ["VAL-B"], depends_on=["b"]),
             _task("ga", "gate", ["VAL-A"], depends_on=["va"]),
             _task("gb", "gate", ["VAL-B"], depends_on=["vb"]),
         ]))
diff --git a/zenith/tests/test_server.py b/zenith/tests/test_server.py
index 4b5533c..437b689 100644
--- a/zenith/tests/test_server.py
+++ b/zenith/tests/test_server.py
@@ -178,12 +178,12 @@ async def test_submit_terminal_review_writes_file(tmp_path: Path, monkeypatch) -
 @pytest.mark.asyncio
 async def test_orchestrator_end_to_end_in_process(
     config: HarnessConfig, workspace: Path
-) -> None:
+    ) -> None:
     def responder(req):
-        if req.node.type == "work":
-            return WorkHandoff(node_id=req.node.id, done=True, report="ok")
+        if req.task.type == "work":
+            return WorkHandoff(node_id=req.task.id, done=True, report="ok")
         return ValidateHandoff(
-            node_id=req.node.id,
+            node_id=req.task.id,
             done=True,
             report="audited",
             items=[ValidationItem(item_id="VAL-001", passed=True)],
@@ -206,8 +206,8 @@ def responder(req):
     (contract_dir / "VAL-001.md").write_text("# VAL-001\n")
     task_list_dict = {
         "tasks": [
-            {"id": "w1", "type": "work", "body": "do", "targets": ["VAL-001"], "skill": "s", "depends_on": []},
-            {"id": "v1", "type": "validate", "body": "audit", "targets": ["VAL-001"], "skill": "aud", "depends_on": ["w1"]},
+            {"id": "w1", "type": "work", "body": "do", "targets": ["VAL-001"], "skill": "engineering-mission-playbook", "depends_on": []},
+            {"id": "v1", "type": "validate", "body": "audit", "targets": ["VAL-001"], "skill": "scrutiny-validator", "depends_on": ["w1"]},
             {"id": "g1", "type": "gate", "body": "", "targets": ["VAL-001"], "skill": None, "depends_on": ["v1"]},
         ],
     }
@@ -281,8 +281,8 @@ async def test_advance_project_tolerates_asyncio_run_dispatcher(
     (contract_dir / "VAL-001.md").write_text("# VAL-001\n")
     task_list_dict = {
         "tasks": [
-            {"id": "w1", "type": "work", "body": "do", "targets": ["VAL-001"], "skill": "s", "depends_on": []},
-            {"id": "v1", "type": "validate", "body": "audit", "targets": ["VAL-001"], "skill": "aud", "depends_on": ["w1"]},
+            {"id": "w1", "type": "work", "body": "do", "targets": ["VAL-001"], "skill": "engineering-mission-playbook", "depends_on": []},
+            {"id": "v1", "type": "validate", "body": "audit", "targets": ["VAL-001"], "skill": "scrutiny-validator", "depends_on": ["w1"]},
             {"id": "g1", "type": "gate", "body": "", "targets": ["VAL-001"], "skill": None, "depends_on": ["v1"]},
         ],
     }
diff --git a/zenith/tests/test_smoke_real_acp.py b/zenith/tests/test_smoke_real_acp.py
index 66b65b6..6d38c79 100644
--- a/zenith/tests/test_smoke_real_acp.py
+++ b/zenith/tests/test_smoke_real_acp.py
@@ -141,14 +141,13 @@ def _build_task_list() -> TaskList:
     ])
 
 
-def _write_contract(workspace: Path, mission_id: str) -> None:
-    d = workspace / ".zenith" / "missions" / mission_id / "contract"
-    d.mkdir(parents=True, exist_ok=True)
+def _write_contract(store: ProjectStore, pid: str, mission_id: str) -> None:
+    d = store.ensure_contract_dir(pid, mission_id)
     (d / "VAL-HELLO-001.md").write_text(CONTRACT_BODY)
 
 
-def _write_worker_skill(workspace: Path) -> None:
-    d = workspace / ".zenith" / "skills" / "hello-file-worker"
+def _write_worker_skill(store: ProjectStore, pid: str) -> None:
+    d = store.zenith_dir(pid) / "skills" / "hello-file-worker"
     d.mkdir(parents=True, exist_ok=True)
     (d / "SKILL.md").write_text(
         """---
@@ -253,12 +252,12 @@ def test_smoke_hello_mission(
 
     # 1) start_project
     start_env = controller.start_project("smoke test brief", str(workspace))
-    pid = ProjectStore(config).list_projects()[0].id
+    pid = controller.store.list_projects()[0].id
     assert start_env.state.state == "mission_planning"
 
     # 2) contract + submit_plan
-    _write_contract(workspace, "mission-001")
-    _write_worker_skill(workspace)
+    _write_contract(controller.store, pid, "mission-001")
+    _write_worker_skill(controller.store, pid)
     plan_env = controller.submit_plan(pid, _build_task_list())
     assert plan_env.state.state == "mission_running"
 
@@ -266,10 +265,12 @@ def test_smoke_hello_mission(
     deadline = time.monotonic() + _TIMEOUT_S
     state_history: list[str] = []
 
-    # Drive: w1 → v1 → g1 → terminal review.
+    # Drive: w1 → v1 → g1, acknowledge the gate checkpoint, then request
+    # closure via end_mission.
     # Each `advance_project` call returns at the next attention/terminal/idle.
     # We expect: first call → gate_checkpoint; we say continue;
-    # second call → Done (via clean terminal reviewer).
+    # second call → mission_running with no runnable task work;
+    # end_mission → Done (via clean terminal reviewer).
     advance_env = controller.advance_project(pid)
     state_history.append(advance_env.state.state)
     assert (
@@ -289,7 +290,13 @@ def test_smoke_hello_mission(
     if time.monotonic() > deadline:
         pytest.fail(f"smoke test exceeded {_TIMEOUT_S}s before terminal review")
 
-    final = controller.advance_project(pid)
+    post_gate = controller.advance_project(pid)
+    state_history.append(post_gate.state.state)
+    assert post_gate.state.state == "mission_running", (
+        f"expected mission_running after gate acknowledgement, got {post_gate.state.state}; "
+        f"history={state_history}"
+    )
+    final = controller.end_mission(pid)
     state_history.append(final.state.state)
     assert final.state.state == "done", (
         f"expected done after terminal review, got {final.state.state}; "
@@ -335,7 +342,7 @@ def test_smoke_hello_mission(
     assert item_passed, f"VAL-HELLO-001 not marked passed: items={val_handoff.items}"
 
     # 7) Closeout written
-    closeout = workspace / ".zenith" / "missions" / "mission-001" / "closeout.md"
+    closeout = controller.store.mission_dir(pid, "mission-001") / "closeout.md"
     assert closeout.exists(), f"missing closeout at {closeout}"
 
 
@@ -377,8 +384,9 @@ def _run_with_real_terminal_reviewer(
     controller = ProjectController(config, dispatcher, reviewer)
 
     controller.start_project("smoke test brief", str(workspace))
-    pid = ProjectStore(config).list_projects()[0].id
-    _write_contract(workspace, "mission-001")
+    pid = controller.store.list_projects()[0].id
+    _write_contract(controller.store, pid, "mission-001")
+    _write_worker_skill(controller.store, pid)
     controller.submit_plan(pid, _build_task_list())
 
     deadline = time.monotonic() + _TIMEOUT_S
@@ -391,14 +399,16 @@ def _run_with_real_terminal_reviewer(
     )
     if time.monotonic() > deadline:
         pytest.fail("smoke test ran out of budget before terminal review")
-    final = controller.advance_project(pid)
+    post_gate = controller.advance_project(pid)
+    assert post_gate.state.state == "mission_running"
+    final = controller.end_mission(pid)
     # Either Done (clean review) or attention_needed (reviewer found gaps).
     # We accept both — the goal is to confirm the L3 ACP wiring works.
     assert final.state.state in ("done", "attention_needed"), (
         f"unexpected post-terminal state: {final.state.state}"
     )
 
-    review_dir = workspace / ".zenith" / "missions" / "mission-001" / "terminal-reviews"
+    review_dir = controller.store.terminal_reviews_dir(pid, "mission-001")
     assert review_dir.is_dir() and any(review_dir.iterdir()), (
         f"terminal reviewer did not write terminal review at {review_dir}"
     )
diff --git a/zenith/tests/test_terminal_review.py b/zenith/tests/test_terminal_review.py
index 4c19166..abf221d 100644
--- a/zenith/tests/test_terminal_review.py
+++ b/zenith/tests/test_terminal_review.py
@@ -49,7 +49,11 @@ def config(harness_home: Path) -> HarnessConfig:
 def _task(tid: str, ttype: str, targets: list[str], skill: str | None = None,
           depends_on: list[str] | None = None) -> Task:
     if skill is None and ttype != "gate":
-        skill = "s"
+        skill = (
+            "scrutiny-validator"
+            if ttype == "validate"
+            else "engineering-mission-playbook"
+        )
     return Task(
         id=tid,
         type=ttype,  # type: ignore[arg-type]
@@ -63,7 +67,7 @@ def _task(tid: str, ttype: str, targets: list[str], skill: str | None = None,
 def _simple_tl() -> TaskList:
     return TaskList(tasks=[
         _task("w1", "work", ["VAL-001"]),
-        _task("v1", "validate", ["VAL-001"], skill="aud", depends_on=["w1"]),
+        _task("v1", "validate", ["VAL-001"], depends_on=["w1"]),
         _task("g1", "gate", ["VAL-001"], depends_on=["v1"]),
     ])