Merge branch 'main' into feature/python-foundry-hosted-agent-vnext

python/packages/core/agent_framework/_skills.py

@@ -486,8 +486,8 @@ YAML_KV_RE = re.compile(
 )
 
 # Validates skill names: lowercase letters, numbers, hyphens only;
-# must not start or end with a hyphen.
-VALID_NAME_RE = re.compile(r"^[a-z0-9]([a-z0-9\-]*[a-z0-9])?$")
+# must not start or end with a hyphen, and must not contain consecutive hyphens.
+VALID_NAME_RE = re.compile(r"^[a-z0-9]([a-z0-9]*-[a-z0-9])*[a-z0-9]*$")
 
 # Default system prompt template for advertising available skills to the model.
 # Use {skills} as the placeholder for the generated skills XML list.
@@ -1156,7 +1156,8 @@ def _validate_skill_metadata(
     if len(name) > MAX_NAME_LENGTH or not VALID_NAME_RE.match(name):
         return (
             f"Skill from '{source}' has an invalid name '{name}': Must be {MAX_NAME_LENGTH} characters or fewer, "
-            "using only lowercase letters, numbers, and hyphens, and must not start or end with a hyphen."
+            "using only lowercase letters, numbers, and hyphens, and must not start or end with a hyphen "
+            "or contain consecutive hyphens."
         )
 
     if not description or not description.strip():
@@ -1241,6 +1242,17 @@ def _read_and_parse_skill_file(
         return None
 
     name, description = result
+
+    dir_name = Path(skill_dir_path).name
+    if name != dir_name:
+        logger.error(
+            "SKILL.md at '%s' has frontmatter name '%s' that does not match the directory name '%s'; skipping.",
+            skill_file,
+            name,
+            dir_name,
+        )
+        return None
+
     return name, description, content
 
 

python/packages/core/agent_framework/_types.py

@@ -2816,6 +2816,7 @@ class ResponseStream(AsyncIterable[UpdateT], Generic[UpdateT, FinalT]):
             cleanup_hooks if cleanup_hooks is not None else []
         )
         self._cleanup_run: bool = False
+        self._stream_error: Exception | None = None
         self._inner_stream: ResponseStream[Any, Any] | None = None
         self._inner_stream_source: ResponseStream[Any, Any] | Awaitable[ResponseStream[Any, Any]] | None = None
         self._wrap_inner: bool = False
@@ -2948,8 +2949,12 @@ class ResponseStream(AsyncIterable[UpdateT], Generic[UpdateT, FinalT]):
             await self._run_cleanup_hooks()
             await self.get_final_response()
             raise
-        except Exception:
-            await self._run_cleanup_hooks()
+        except Exception as exc:
+            self._stream_error = exc
+            try:
+                await self._run_cleanup_hooks()
+            finally:
+                self._stream_error = None
             raise
         if self._map_update is not None:
             update = self._map_update(update)  # type: ignore[assignment]

python/packages/core/agent_framework/_workflows/_agent.py

@@ -119,15 +119,11 @@ class WorkflowAgent(BaseAgent):
         if not any(is_type_compatible(list[Message], input_type) for input_type in start_executor.input_types):
             raise ValueError("Workflow's start executor cannot handle list[Message]")
 
-        resolved_context_providers = list(context_providers) if context_providers is not None else []
-        if not resolved_context_providers:
-            resolved_context_providers.append(InMemoryHistoryProvider())
-
         super().__init__(
             id=id,
             name=name,
             description=description,
-            context_providers=resolved_context_providers,
+            context_providers=context_providers,
             **kwargs,
         )
         self._workflow: Workflow = workflow
@@ -261,6 +257,15 @@ class WorkflowAgent(BaseAgent):
             An AgentResponse representing the workflow execution results.
         """
         input_messages = normalize_messages_input(messages)
+
+        if (
+            not any(
+                provider.load_messages for provider in self.context_providers if isinstance(provider, HistoryProvider)
+            )
+            and session is not None
+        ):
+            self.context_providers.append(InMemoryHistoryProvider())
+
         provider_session = session
         if provider_session is None and self.context_providers:
             provider_session = AgentSession()
@@ -332,6 +337,15 @@ class WorkflowAgent(BaseAgent):
             AgentResponseUpdate objects representing the workflow execution progress.
         """
         input_messages = normalize_messages_input(messages)
+
+        if (
+            not any(
+                provider.load_messages for provider in self.context_providers if isinstance(provider, HistoryProvider)
+            )
+            and session is not None
+        ):
+            self.context_providers.append(InMemoryHistoryProvider())
+
         provider_session = session
         if provider_session is None and self.context_providers:
             provider_session = AgentSession()

python/packages/core/agent_framework/ag_ui/__init__.py

@@ -7,10 +7,13 @@ This module lazily re-exports objects from:
 
 Supported classes and functions:
 - AgentFrameworkAgent
+- AgentFrameworkWorkflow
 - AGUIChatClient
 - AGUIEventConverter
 - AGUIHttpService
 - add_agent_framework_fastapi_endpoint
+- state_update
+- __version__
 """
 
 import importlib
@@ -23,6 +26,10 @@ _IMPORTS = [
     "AgentFrameworkWorkflow",
     "add_agent_framework_fastapi_endpoint",
     "AGUIChatClient",
+    "AGUIEventConverter",
+    "AGUIHttpService",
+    "state_update",
+    "__version__",
 ]
 
 

python/packages/core/agent_framework/ag_ui/__init__.pyi

@@ -8,6 +8,7 @@ from agent_framework_ag_ui import (
     AGUIHttpService,
     __version__,
     add_agent_framework_fastapi_endpoint,
+    state_update,
 )
 
 __all__ = [
@@ -18,4 +19,5 @@ __all__ = [
     "AgentFrameworkWorkflow",
     "__version__",
     "add_agent_framework_fastapi_endpoint",
+    "state_update",
 ]

python/packages/core/agent_framework/observability.py

@@ -1323,6 +1323,12 @@ class ChatTelemetryLayer(Generic[OptionsCoT]):
                 from ._types import ChatResponse
 
                 try:
+                    if result_stream._stream_error is not None:  # pyright: ignore[reportPrivateUsage]
+                        # Stream errored; skip get_final_response() to avoid firing
+                        # result hooks such as after_run context providers on error
+                        # paths. Capture the error on the span before returning.
+                        capture_exception(span=span, exception=result_stream._stream_error, timestamp=time_ns())  # pyright: ignore[reportPrivateUsage]
+                        return
                     response: ChatResponse[Any] = await result_stream.get_final_response()
                     duration = duration_state.get("duration")
                     response_attributes = _get_response_attributes(attributes, response)
@@ -1579,6 +1585,12 @@ class AgentTelemetryLayer:
                 from ._types import AgentResponse
 
                 try:
+                    if result_stream._stream_error is not None:  # pyright: ignore[reportPrivateUsage]
+                        # Stream errored; skip get_final_response() to avoid firing
+                        # result hooks such as after_run context providers on error
+                        # paths. Capture the error on the span before returning.
+                        capture_exception(span=span, exception=result_stream._stream_error, timestamp=time_ns())  # pyright: ignore[reportPrivateUsage]
+                        return
                     response: AgentResponse[Any] = await result_stream.get_final_response()
                     duration = duration_state.get("duration")
                     response_attributes = _get_response_attributes(

python/packages/core/pyproject.toml

@@ -34,14 +34,13 @@ all = [
     "mcp>=1.24.0,<2",
     "agent-framework-a2a",
     "agent-framework-ag-ui",
+    "agent-framework-anthropic",
     "agent-framework-azure-ai-search",
     "agent-framework-azure-cosmos",
-    "agent-framework-anthropic",
-    "agent-framework-openai",
-    "agent-framework-claude",
     "agent-framework-azurefunctions",
     "agent-framework-bedrock",
     "agent-framework-chatkit",
+    "agent-framework-claude",
     "agent-framework-copilotstudio",
     "agent-framework-declarative",
     "agent-framework-devui",
@@ -52,6 +51,7 @@ all = [
     "agent-framework-lab",
     "agent-framework-mem0",
     "agent-framework-ollama",
+    "agent-framework-openai",
     "agent-framework-orchestrations",
     "agent-framework-purview",
     "agent-framework-redis",

python/packages/core/tests/core/test_skills.py

@@ -296,6 +296,15 @@ class TestDiscoverAndLoadSkills:
         skills = _discover_file_skills([str(tmp_path)])
         assert len(skills) == 0
 
+    def test_skips_skill_with_name_directory_mismatch(self, tmp_path: Path) -> None:
+        skill_dir = tmp_path / "wrong-dir-name"
+        skill_dir.mkdir()
+        (skill_dir / "SKILL.md").write_text(
+            "---\nname: actual-skill-name\ndescription: A skill.\n---\nBody.", encoding="utf-8"
+        )
+        skills = _discover_file_skills([str(tmp_path)])
+        assert len(skills) == 0
+
     def test_deduplicates_skill_names(self, tmp_path: Path) -> None:
         dir1 = tmp_path / "dir1"
         dir2 = tmp_path / "dir2"
@@ -904,6 +913,11 @@ class TestSkill:
         provider = SkillsProvider(skills=[invalid_skill])
         assert len(provider._skills) == 0
 
+    def test_name_with_consecutive_hyphens_skipped(self) -> None:
+        invalid_skill = Skill(name="consecutive--hyphens", description="A skill.", content="Body")
+        provider = SkillsProvider(skills=[invalid_skill])
+        assert len(provider._skills) == 0
+
     def test_name_too_long_skipped(self) -> None:
         invalid_skill = Skill(name="a" * 65, description="A skill.", content="Body")
         provider = SkillsProvider(skills=[invalid_skill])
@@ -1421,6 +1435,11 @@ class TestValidateSkillMetadata:
         assert result is not None
         assert "invalid name" in result
 
+    def test_name_with_consecutive_hyphens(self) -> None:
+        result = _validate_skill_metadata("consecutive--hyphens", "desc", "source")
+        assert result is not None
+        assert "invalid name" in result
+
     def test_single_char_name(self) -> None:
         assert _validate_skill_metadata("a", "desc", "source") is None
 
@@ -1526,6 +1545,15 @@ class TestReadAndParseSkillFile:
         result = _read_and_parse_skill_file(str(skill_dir))
         assert result is None
 
+    def test_name_directory_mismatch_returns_none(self, tmp_path: Path) -> None:
+        skill_dir = tmp_path / "wrong-dir-name"
+        skill_dir.mkdir()
+        (skill_dir / "SKILL.md").write_text(
+            "---\nname: actual-skill-name\ndescription: A skill.\n---\nBody.", encoding="utf-8"
+        )
+        result = _read_and_parse_skill_file(str(skill_dir))
+        assert result is None
+
 
 # ---------------------------------------------------------------------------
 # Tests: _create_resource_element

python/packages/core/tests/workflow/test_workflow_agent.py

@@ -14,6 +14,7 @@ from agent_framework import (
     AgentSession,
     Content,
     Executor,
+    HistoryProvider,
     InMemoryHistoryProvider,
     Message,
     ResponseStream,
@@ -678,6 +679,110 @@ class TestWorkflowAgent:
 
         assert agent.context_providers == [explicit_provider]
 
+    async def test_no_history_provider_injected_when_session_is_none(self) -> None:
+        """Test that InMemoryHistoryProvider is NOT injected when session is None."""
+        capturing_executor = ConversationHistoryCapturingExecutor(id="no_session_test")
+        workflow = WorkflowBuilder(start_executor=capturing_executor).build()
+        agent = WorkflowAgent(workflow=workflow, name="No Session Agent")
+
+        await agent.run("hello")
+
+        assert not any(isinstance(p, InMemoryHistoryProvider) for p in agent.context_providers)
+
+    async def test_no_history_provider_injected_when_session_is_none_streaming(self) -> None:
+        """Test that InMemoryHistoryProvider is NOT injected when session is None (streaming)."""
+        capturing_executor = ConversationHistoryCapturingExecutor(id="no_session_stream_test")
+        workflow = WorkflowBuilder(start_executor=capturing_executor).build()
+        agent = WorkflowAgent(workflow=workflow, name="No Session Stream Agent")
+
+        async for _ in agent.run("hello", stream=True):
+            pass
+
+        assert not any(isinstance(p, InMemoryHistoryProvider) for p in agent.context_providers)
+
+    async def test_no_injection_when_history_provider_with_load_messages_exists(self) -> None:
+        """Test that no InMemoryHistoryProvider is injected when an existing HistoryProvider has load_messages=True."""
+        capturing_executor = ConversationHistoryCapturingExecutor(id="existing_provider_test")
+        workflow = WorkflowBuilder(start_executor=capturing_executor).build()
+        existing_provider = InMemoryHistoryProvider("custom", load_messages=True)
+        agent = WorkflowAgent(
+            workflow=workflow,
+            name="Existing Provider Agent",
+            context_providers=[existing_provider],
+        )
+        session = AgentSession()
+
+        await agent.run("hello", session=session)
+
+        # Should still have only the original provider
+        history_providers = [p for p in agent.context_providers if isinstance(p, HistoryProvider)]
+        assert len(history_providers) == 1
+        assert history_providers[0] is existing_provider
+
+    async def test_injection_when_history_provider_with_load_messages_false(self) -> None:
+        """Test that InMemoryHistoryProvider IS injected when existing HistoryProvider has load_messages=False."""
+        capturing_executor = ConversationHistoryCapturingExecutor(id="no_load_provider_test")
+        workflow = WorkflowBuilder(start_executor=capturing_executor).build()
+        audit_provider = InMemoryHistoryProvider("audit", load_messages=False)
+        agent = WorkflowAgent(
+            workflow=workflow,
+            name="Audit Provider Agent",
+            context_providers=[audit_provider],
+        )
+        session = AgentSession()
+
+        await agent.run("hello", session=session)
+
+        # Should have injected an additional InMemoryHistoryProvider with load_messages=True
+        history_providers = [p for p in agent.context_providers if isinstance(p, HistoryProvider)]
+        assert len(history_providers) == 2
+        loading_providers = [p for p in history_providers if p.load_messages]
+        assert len(loading_providers) == 1
+        assert isinstance(loading_providers[0], InMemoryHistoryProvider)
+
+    async def test_no_duplicate_injection_on_multiple_runs(self) -> None:
+        """Test that calling run() multiple times does not keep adding InMemoryHistoryProvider."""
+        capturing_executor = ConversationHistoryCapturingExecutor(id="no_dup_test")
+        workflow = WorkflowBuilder(start_executor=capturing_executor).build()
+        agent = WorkflowAgent(workflow=workflow, name="No Dup Agent")
+        session = AgentSession()
+
+        await agent.run("first", session=session)
+        await agent.run("second", session=session)
+        await agent.run("third", session=session)
+
+        history_providers = [p for p in agent.context_providers if isinstance(p, InMemoryHistoryProvider)]
+        assert len(history_providers) == 1
+
+    async def test_no_duplicate_injection_on_multiple_runs_streaming(self) -> None:
+        """Test that calling run(stream=True) multiple times does not keep adding InMemoryHistoryProvider."""
+        capturing_executor = ConversationHistoryCapturingExecutor(id="no_dup_stream_test")
+        workflow = WorkflowBuilder(start_executor=capturing_executor).build()
+        agent = WorkflowAgent(workflow=workflow, name="No Dup Stream Agent")
+        session = AgentSession()
+
+        async for _ in agent.run("first", stream=True, session=session):
+            pass
+        async for _ in agent.run("second", stream=True, session=session):
+            pass
+        async for _ in agent.run("third", stream=True, session=session):
+            pass
+
+        history_providers = [p for p in agent.context_providers if isinstance(p, InMemoryHistoryProvider)]
+        assert len(history_providers) == 1
+
+    async def test_injection_with_session_in_streaming_mode(self) -> None:
+        """Test that InMemoryHistoryProvider is injected when session is provided in streaming mode."""
+        capturing_executor = ConversationHistoryCapturingExecutor(id="stream_inject_test")
+        workflow = WorkflowBuilder(start_executor=capturing_executor).build()
+        agent = WorkflowAgent(workflow=workflow, name="Stream Inject Agent")
+        session = AgentSession()
+
+        async for _ in agent.run("hello", stream=True, session=session):
+            pass
+
+        assert any(isinstance(p, InMemoryHistoryProvider) for p in agent.context_providers)
+
     async def test_checkpoint_storage_passed_to_workflow(self) -> None:
         """Test that checkpoint_storage parameter is passed through to the workflow."""
         from agent_framework import InMemoryCheckpointStorage