Fix tests for fake LLM response

tests/test_agent.py

@@ -1,13 +1,14 @@
-from agt.agent import Agent, Tool, echo_tool
+from agt.agent import Agent, FakeLLMClient, Tool, echo_tool
 
 
-def test_agent_returns_default_message() -> None:
-    agent = Agent()
+def test_agent_returns_fake_llm_message() -> None:
+    agent = Agent(llm=FakeLLMClient())
     agent.register_tool(Tool("echo", "Echo input text", echo_tool))
 
     result = agent.run("hello")
 
-    assert "教学用 Agent 骨架" in result
+    assert "[fake-llm]" in result
+    assert "hello" in result
     assert agent.messages[0].role == "user"
     assert agent.messages[-1].role == "assistant"
 
@@ -21,3 +22,13 @@ def test_load_skills_finds_skill_md(tmp_path) -> None:
     skills = agent.load_skills(tmp_path / "skills")
 
     assert skills == ["writing"]
+
+
+def test_fake_llm_streams_multiple_chunks() -> None:
+    llm = FakeLLMClient()
+    from agt.agent import Message
+
+    chunks = list(llm.stream_text([Message(role="user", content="x")]))
+
+    assert len(chunks) >= 2
+    assert "".join(chunks).startswith("[fake-llm]")