feat(engine): 添加运行时上下文支持并重构工具迭代限制

添加 RuntimeContext 类用于捕获模型运行时的日期时间信息，
包括UTC时间、本地时间和时区信息，并在系统提示中显示这些信息。

同时增加最大上下文消息数和工具迭代次数的配置选项，
将验证服务从引擎加载器中移除，并更新相关的数据结构和接口。

BREAKING CHANGE: 移除了验证服务，相关字段被替换为证据状态和接受状态。

- 添加 RuntimeContext 类和相关渲染方法
- 增加 max_context_messages 和 max_tool_iterations 配置
- 移除 ValidationService 相关代码
- 更新消息记录中的验证状态字段
- 添加原始工具调用检测和回退处理

app-instance/backend/tests/unit/test_phase5_skills_runtime.py

@@ -35,6 +35,7 @@ class StubProvider(LLMProvider):
         model: str | None = None,
         max_tokens: int = 4096,
         temperature: float = 0.7,
+        thinking_enabled: bool | None = None,
     ) -> LLMResponse:
         if not self._responses:
             raise AssertionError("No stubbed provider responses left")
@@ -47,11 +48,22 @@ class StubProvider(LLMProvider):
 class StubSkillAssembler:
     def __init__(self, activated_skills: list[SkillContext]) -> None:
         self.activated_skills = activated_skills
+        self.calls: list[dict] = []
 
     async def assemble(self, **kwargs) -> SkillAssemblyResult:
+        self.calls.append(kwargs)
         return SkillAssemblyResult(activated_skills=list(self.activated_skills))
 
 
+class RecordingToolAssembler:
+    def __init__(self) -> None:
+        self.calls: list[dict] = []
+
+    async def assemble(self, **kwargs):
+        self.calls.append(kwargs)
+        return kwargs["registry"].get_specs(["memory"])
+
+
 def _tool_call(*, name: str = "echo", arguments: dict | None = None, call_id: str = "call-1") -> SimpleNamespace:
     return SimpleNamespace(
         id=call_id,
@@ -576,6 +588,48 @@ def test_agent_loop_records_skill_receipts_and_effects(tmp_path: Path) -> None:
     assert effect_records[-1].run_id == result.run_id
 
 
+def test_thinking_disabled_still_uses_skill_and_tool_assembly(tmp_path: Path) -> None:
+    skill = SkillContext(
+        name="docker-debug",
+        content="Use docker logs before editing config.",
+        version="v0007",
+        content_hash="hash-v7",
+        activation_reason="llm_selected",
+        tool_hints=["terminal"],
+    )
+    skill_assembler = StubSkillAssembler([skill])
+    tool_assembler = RecordingToolAssembler()
+    loader = EngineLoader(
+        workspace=tmp_path,
+        skill_assembler=skill_assembler,
+        tool_assembler=tool_assembler,
+    )
+    loop = AgentLoop(loader=loader)
+    bundle = ProviderBundle(
+        main_runtime=SimpleNamespace(model="stub-model", provider_name="stub"),
+        main_provider=StubProvider(
+            [LLMResponse(content="Done", finish_reason="stop", provider_name="stub", model="stub-model")]
+        ),
+    )
+
+    result = asyncio.run(
+        loop.process_direct(
+            "Why is the Docker container crashing?",
+            provider_bundle=bundle,
+            thinking_enabled=False,
+        )
+    )
+    loaded = loop.boot()
+    events = loaded.session_manager.get_run_event_records(result.session_id, result.run_id)
+    tool_selection = next(event for event in events if event.event_type == "tool_selection_snapshotted")
+
+    assert skill_assembler.calls
+    assert skill_assembler.calls[0]["thinking_enabled"] is False
+    assert tool_assembler.calls
+    assert [skill.name for skill in tool_assembler.calls[0]["activated_skills"]] == ["docker-debug"]
+    assert tool_selection.event_payload["tool_names"] == ["memory"]
+
+
 def test_agent_loop_records_max_tool_iterations_as_failed_skill_effect(tmp_path: Path) -> None:
     skill = SkillContext(
         name="docker-debug",
@@ -635,6 +689,52 @@ def test_agent_loop_records_max_tool_iterations_as_failed_skill_effect(tmp_path:
     assert effect_records[-1].success is False
 
 
+def test_agent_loop_suppresses_raw_tool_call_when_finalizing_after_tool_limit(tmp_path: Path) -> None:
+    loader = EngineLoader(
+        workspace=tmp_path,
+        skill_assembler=StubSkillAssembler([]),
+    )
+    loop = AgentLoop(loader=loader)
+    bundle = ProviderBundle(
+        main_runtime=SimpleNamespace(model="stub-model", provider_name="stub"),
+        main_provider=StubProvider(
+            [
+                LLMResponse(
+                    content="Need a tool.",
+                    finish_reason="tool_calls",
+                    tool_calls=[_tool_call()],
+                    provider_name="stub",
+                    model="stub-model",
+                ),
+                LLMResponse(
+                    content=(
+                        "<tool_call>\n"
+                        "<function=mcp_local_web_mcp_web_fetch>\n"
+                        "<parameter=url>https://example.com</parameter>\n"
+                        "</function>\n"
+                        "</tool_call>"
+                    ),
+                    finish_reason="stop",
+                    provider_name="stub",
+                    model="stub-model",
+                ),
+            ]
+        ),
+    )
+
+    result = asyncio.run(
+        loop.process_direct(
+            "Fetch the latest result",
+            provider_bundle=bundle,
+            max_tool_iterations=0,
+        )
+    )
+
+    assert result.finish_reason == "max_tool_iterations"
+    assert "<tool_call>" not in result.output_text
+    assert "raw tool call was suppressed" in result.output_text
+
+
 def test_llm_request_snapshot_defaults_to_compact_payload(tmp_path: Path) -> None:
     loop = AgentLoop(loader=EngineLoader(workspace=tmp_path, skill_assembler=StubSkillAssembler([])))
     bundle = ProviderBundle(