diff --git a/.github/workflows/tests.yml b/.github/workflows/tests.yml
new file mode 100644
index 00000000..b569087f
--- /dev/null
+++ b/.github/workflows/tests.yml
@@ -0,0 +1,44 @@
+name: 🧪 Tests
+
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+concurrency:
+  group: tests-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  pytest:
+    name: 🐍 ${{ matrix.os }} · py${{ matrix.python }}
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        # Ubuntu only while this branch is based on a main that predates the
+        # Windows support in #57: the Windows lanes there exercise platform
+        # code this branch does not carry, and fail on main's known
+        # POSIX-only layers (os.killpg and friends). #57 brings the full
+        # ubuntu + windows matrix; on merge, its version of this file wins.
+        os: [ubuntu-latest]
+        python: ["3.10", "3.14"]
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python }}
+
+      - name: 📦 Install with test dependencies
+        run: python -m pip install -e ".[test]"
+
+      - name: 🧪 Run the suite
+        # The Windows symlink fixtures skip themselves when the runner lacks
+        # SeCreateSymbolicLinkPrivilege; those skips are expected and green.
+        run: python -m pytest -q
diff --git a/src/lh_harness/role_prompts.py b/src/lh_harness/role_prompts.py
index d3b38847..03c419ef 100644
--- a/src/lh_harness/role_prompts.py
+++ b/src/lh_harness/role_prompts.py
@@ -480,12 +480,21 @@ def format_audit_findings(
 
 def parse_role_manager_next_step(text: str) -> RoleNextStep:
     for line in str(text or "").splitlines():
-        normalized = line.strip().strip("*").replace(" ", "").replace("　", "").lower()
+        # `*` covers bold; the backtick covers code spans. The prompt itself
+        # displays every route inside backticks ("exactly one route:
+        # `Next: gui`, ..."), and models -- reliably the smaller ones -- copy
+        # that formatting into their answer, so a parser that accepts
+        # `**Next: cli**` but not "`Next: cli`" burns a whole round on
+        # formatting the harness's own instruction taught the model.
+        normalized = line.strip().strip("*`").replace(" ", "").replace("　", "").lower()
         # Models commonly append a short rationale after the required route,
         # e.g. `Next: done — all constraints passed`. Treat only an explicitly
         # delimited suffix as commentary so prose such as `Next: done later`
         # remains invalid.
-        normalized = re.split(r"(?:—|–|--|//|#|[（(])", normalized, maxsplit=1)[0]
+        # Strip wrappers again after cutting the rationale: in
+        # "`Next: cli` — reason" the closing backtick sits before the dash and
+        # survives the first strip.
+        normalized = re.split(r"(?:—|–|--|//|#|[（(])", normalized, maxsplit=1)[0].strip("*`")
         if normalized in {"下一步:gui任务", "下一步：gui任务", "next:gui"}:
             return MANAGER_NEXT_GUI
         if normalized in {"下一步:cli任务", "下一步：cli任务", "next:cli"}:
diff --git a/tests/test_role_prompts.py b/tests/test_role_prompts.py
index 290980e1..68cf3346 100644
--- a/tests/test_role_prompts.py
+++ b/tests/test_role_prompts.py
@@ -94,6 +94,21 @@ def test_manager_route_rejects_undelimited_suffix() -> None:
     assert parse_role_manager_next_step("Next: done later") != MANAGER_NEXT_DONE
 
 
+@pytest.mark.parametrize(
+    "line",
+    [
+        # The prompt shows every route inside backticks, and models copy that
+        # formatting literally; bold was already accepted, so a code span must
+        # not be the one markdown wrapper that voids an otherwise valid route.
+        "`Next: cli`",
+        "**`Next: cli`**",
+        "`Next: cli` — routed to the CLI executor",
+    ],
+)
+def test_manager_route_accepts_code_span_wrapping(line: str) -> None:
+    assert parse_role_manager_next_step(line) == MANAGER_NEXT_CLI
+
+
 @pytest.mark.parametrize("language", ["en", "zh"])
 def test_manager_prompt_exposes_total_and_remaining_round_budget(language: str) -> None:
     prompt = build_role_manager_prompt(