feat(task): synthesize and validate from evidence

app-instance/backend/tests/unit/test_task_mode_feedback.py

@@ -20,7 +20,7 @@ class StubProvider(LLMProvider):
     def __init__(self, responses: list[LLMResponse]) -> None:
         super().__init__()
         self._responses = list(responses)
-        self.calls: list[list[dict]] = []
+        self.calls: list[dict[str, object]] = []
 
     async def chat(
         self,
@@ -30,7 +30,7 @@ class StubProvider(LLMProvider):
         max_tokens: int = 4096,
         temperature: float = 0.7,
     ) -> LLMResponse:
-        self.calls.append(messages)
+        self.calls.append({"messages": messages, "tools": tools, "model": model})
         if not self._responses:
             raise AssertionError("No stubbed provider responses left")
         return self._responses.pop(0)
@@ -42,8 +42,10 @@ class StubProvider(LLMProvider):
 class StubValidationService:
     def __init__(self, results: list[ValidationResult]) -> None:
         self.results = list(results)
+        self.calls: list[dict] = []
 
     async def validate_task_result(self, **kwargs) -> ValidationResult:
+        self.calls.append(kwargs)
         if not self.results:
             raise AssertionError("No stubbed validation results left")
         return self.results.pop(0)
@@ -706,10 +708,45 @@ def test_task_mode_team_plan_runs_subagent_then_main_synthesis(tmp_path: Path) -
     assert result.run_id == task.run_ids[-1]
     assert any(event.event_type == "task_execution_planned" for event in events)
     assert any(event.event_type == "task_team_run_completed" for event in events)
-    assert "sub-agent evidence" in main_provider.calls[0][0]["content"]
+    assert "sub-agent evidence" in main_provider.calls[0]["messages"][0]["content"]
     assert "sub-agent evidence" != result.output_text
 
 
+def test_task_mode_team_synthesis_runs_without_tools_and_receives_evidence(tmp_path: Path) -> None:
+    main_provider = StubProvider(
+        [
+            LLMResponse(content="final synthesized answer", finish_reason="stop", provider_name="stub", model="stub-model")
+        ]
+    )
+    sub_provider = StubProvider(
+        [
+            LLMResponse(content="sub-agent evidence", finish_reason="stop", provider_name="stub", model="stub-model")
+        ]
+    )
+    validation = StubValidationService([ValidationResult(status="accepted", score=0.9, validator="test")])
+    service = AgentService(
+        loader=EngineLoader(
+            workspace=tmp_path,
+            task_execution_planner=StubTaskExecutionPlanner([_team_plan()]),
+            validation_service=validation,
+        )
+    )
+
+    result = asyncio.run(
+        service.process_direct(
+            "implement team-backed workflow",
+            session_id="web:team-no-tools",
+            provider_bundle=_provider_bundle(main_provider),
+            team_provider_bundle_factory=lambda node: _provider_bundle(sub_provider),
+        )
+    )
+
+    assert result.output_text == "final synthesized answer"
+    assert main_provider.calls[0]["tools"] is None
+    assert "sub-agent evidence" in main_provider.calls[0]["messages"][0]["content"]
+    assert "Task evidence packet" in validation.calls[0]["evidence_text"]
+
+
 def test_task_mode_team_failure_still_uses_main_synthesis(tmp_path: Path) -> None:
     main_provider = StubProvider(
         [
@@ -737,9 +774,9 @@ def test_task_mode_team_failure_still_uses_main_synthesis(tmp_path: Path) -> Non
 
     assert result.output_text == "fallback synthesized answer"
     assert any(event.event_type == "task_team_run_failed" for event in events)
-    assert "sub-agent unavailable" in main_provider.calls[0][0]["content"]
-    assert "same class of tools fails repeatedly" in main_provider.calls[0][0]["content"]
-    assert "user-visible fallback answer" in main_provider.calls[0][0]["content"]
+    assert "sub-agent unavailable" in main_provider.calls[0]["messages"][0]["content"]
+    assert "same class of tools fails repeatedly" in main_provider.calls[0]["messages"][0]["content"]
+    assert "user-visible fallback answer" in main_provider.calls[0]["messages"][0]["content"]
 
 
 def test_task_mode_team_retry_hides_first_synthesis_run(tmp_path: Path) -> None: