fix(api): carry orchestration_markers in the task response

TaskResponse declares the field, but task_to_response never assigned it,
so every GET /tasks response served null. The orchestrator fetches its
work over that endpoint and builds prompts from the returned dict, so
all twelve marker reads in orchestrator.py resolved to `{}`.

Live symptom: a Barfly explorer was told "SCREENED CANDIDATES: (none)"
and correctly refused to invent a tweet — while BarflyEngine only opens
an exploration task when it HAS candidates, so the ones it gathered were
sitting on the row the whole time. The gateway path reads markers off the
ORM object directly, which is why propose_conversation_replies could see
what the prompt could not.

Same blindness applied to every other marker consumer on the dispatch
path: the spotlight's seen-features dedup, pest control's evidence
snapshot, coroner's autopsy subject, and the dev-redispatch's
original_developer lookup.

tests/unit/api/test_schemas_tasks.py

@@ -42,7 +42,7 @@ from roboco.models.base import (
     Team,
 )
 from roboco.models.product import ProductCellMapping
-from roboco.runtime.orchestrator import AgentOrchestrator
+from roboco.runtime.orchestrator import AgentOrchestrator, _format_barfly_candidates
 
 _ORDER_DEFAULT = 0
 
@@ -312,7 +312,9 @@ def test_task_update_budget_usd_rejects_negative() -> None:
 # ---------------------------------------------------------------------------
 
 
-def _stub_task(*, with_project: bool = False) -> Any:
+def _stub_task(
+    *, with_project: bool = False, markers: dict[str, Any] | None = None
+) -> Any:
     """Build a TaskTable stand-in that matches task_to_response's reads."""
     return SimpleNamespace(
         id=uuid4(),
@@ -360,6 +362,7 @@ def _stub_task(*, with_project: bool = False) -> Any:
         doc_notes=None,
         quick_context=None,
         notes_structured=None,
+        orchestration_markers=markers,
         self_verified=False,
         qa_verified=None,
         branch_name=None,
@@ -368,6 +371,44 @@ def _stub_task(*, with_project: bool = False) -> Any:
     )
 
 
+def _response_for(stub: Any) -> Any:
+    fake_inspector = MagicMock()
+    fake_inspector.unloaded = {"project"}
+    with patch("roboco.api.schemas.tasks.sa_inspect", return_value=fake_inspector):
+        return task_to_response(stub)
+
+
+def test_task_to_response_carries_orchestration_markers() -> None:
+    """The field is declared on TaskResponse, so omitting it from the builder
+    served null instead of failing loudly."""
+    stub = _stub_task(markers={"barfly_candidates": [{"id": "1"}]})
+    assert _response_for(stub).orchestration_markers == {
+        "barfly_candidates": [{"id": "1"}]
+    }
+
+
+def test_task_to_response_markers_reach_a_board_program_prompt() -> None:
+    """The regression this closes end-to-end: the orchestrator fetches work
+    over GET /tasks and renders prompts from that dict, so a dropped marker
+    showed a Barfly explorer "(none)" for candidates its engine HAD gathered
+    — and since the engine opens no task at all when the candidate list is
+    empty, the explorer could only ever conclude there was nothing to do."""
+    candidate = {
+        "id": "1955",
+        "author_handle": "someone",
+        "text": "how do you keep AI agents from stepping on each other?",
+        "engagement_note": "42 likes",
+    }
+    stub = _stub_task(markers={"barfly_candidates": [candidate]})
+    dispatched = _response_for(stub).model_dump()
+
+    rendered = _format_barfly_candidates(dispatched["orchestration_markers"] or {})
+
+    assert rendered != "(none)"
+    assert "id=1955" in rendered
+    assert "how do you keep AI agents" in rendered
+
+
 def test_task_to_response_omits_slug_when_project_not_loaded() -> None:
     stub = _stub_task(with_project=False)
     fake_inspector = MagicMock()