feat(app): 移除内置agents并添加CORS支持和技能上传优化

移除了agents/registry.json中的所有内置agents配置，将agents数组清空。
为web应用添加了CORS中间件支持，允许指定的前端地址跨域访问。
重构了技能上传功能，增加了LLM重写机制，自动规范化上传的技能格式。
新增了工具名称提取逻辑，从技能正文中自动识别Required Tools段落。
更新了技能学习候选者和草稿的载荷结构，添加评估报告统计信息。
修改了意图路由技能的说明，改进任务状态管理逻辑。

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

@@ -19,8 +19,22 @@ from beaver.skills.specs import SkillSpecStore
 
 
 class StubProvider(LLMProvider):
-    async def chat(self, messages: list[dict], tools: list[dict] | None = None, model: str | None = None, max_tokens: int = 4096, temperature: float = 0.7) -> LLMResponse:
-        return LLMResponse(content="ok")
+    def __init__(self, content: str = "ok") -> None:
+        super().__init__()
+        self.content = content
+        self.calls: list[dict] = []
+
+    async def chat(
+        self,
+        messages: list[dict],
+        tools: list[dict] | None = None,
+        model: str | None = None,
+        max_tokens: int = 4096,
+        temperature: float = 0.7,
+        thinking_enabled: bool | None = None,
+    ) -> LLMResponse:
+        self.calls.append({"messages": messages, "model": model, "max_tokens": max_tokens, "temperature": temperature})
+        return LLMResponse(content=self.content)
 
     def get_default_model(self) -> str:
         return "stub"
@@ -92,7 +106,6 @@ def test_eval_pass_allows_publish_after_safety_and_review(tmp_path: Path) -> Non
     report = asyncio.run(pipeline.evaluate_draft("candidate-1", draft.skill_name, draft.draft_id, provider_bundle=_bundle()))
     safety = pipeline.check_safety(draft.skill_name, draft.draft_id)
     pipeline.submit_review(draft.skill_name, draft.draft_id, requested_by="tester")
-    pipeline.approve(draft.skill_name, draft.draft_id, reviewer="tester")
     published = pipeline.publish(draft.skill_name, draft.draft_id, publisher="tester")
 
     assert report.passed is True
@@ -114,7 +127,6 @@ def test_eval_regression_blocks_publish(tmp_path: Path) -> None:
     report = asyncio.run(pipeline.evaluate_draft("candidate-1", draft.skill_name, draft.draft_id, provider_bundle=_bundle()))
     pipeline.check_safety(draft.skill_name, draft.draft_id)
     pipeline.submit_review(draft.skill_name, draft.draft_id, requested_by="tester")
-    pipeline.approve(draft.skill_name, draft.draft_id, reviewer="tester")
 
     assert report.passed is False
     assert pipeline.get_candidate("candidate-1").status == "eval_failed"
@@ -160,7 +172,14 @@ def test_eval_does_not_clear_safety_failed_status(tmp_path: Path) -> None:
 
 
 class FakeReplayRunner:
+    def __init__(self, *, baseline_answer: str = "done", candidate_answer: str = "done") -> None:
+        self.baseline_answer = baseline_answer
+        self.candidate_answer = candidate_answer
+        self.requests = []
+
     async def run_arm(self, request):
+        self.requests.append(request)
+        final_answer = self.candidate_answer if request.arm == "candidate" else self.baseline_answer
         return {
             "case_id": request.case_id,
             "arm": request.arm,
@@ -168,7 +187,7 @@ class FakeReplayRunner:
             "run_id": f"{request.arm}-run",
             "task_text": request.task_text,
             "finish_reason": "stop",
-            "final_answer": "done",
+            "final_answer": final_answer,
             "tool_calls": [
                 {
                     "tool_name": "write_file",
@@ -213,3 +232,102 @@ def test_eval_report_includes_replay_case_and_coverage(tmp_path: Path) -> None:
     assert 0.0 <= report.execution_coverage <= 1.0
     assert 0.0 <= report.surrogate_coverage <= 1.0
     assert report.confidence in {"low", "medium", "high"}
+    assert "ability_score" in report.case_reports[0]
+    assert "tool_execution_score" in report.case_reports[0]
+    assert report.ability_score_summary["score_role"] == "primary"
+    assert report.tool_execution_summary["score_role"] == "diagnostic_only"
+
+
+def test_replay_main_score_uses_validator_not_tool_success(tmp_path: Path) -> None:
+    pipeline = _pipeline(tmp_path)
+    pipeline.learning_store.update_learning_candidate(
+        "candidate-1",
+        evidence={
+            "eval_cases": [
+                {
+                    "run_id": "validator-case",
+                    "task_id": "validator-case",
+                    "session_id": "eval",
+                    "task_text": "Write the release verdict.",
+                    "validator": {
+                        "type": "final_answer_contains",
+                        "required_terms": ["ship"],
+                        "forbidden_terms": ["do not ship"],
+                    },
+                    "accepted_score": 0.5,
+                }
+            ]
+        },
+    )
+    draft = pipeline.draft_service.create_new_skill_draft(
+        skill_name="release-checklist",
+        proposed_content="# Release\n\nRun tests.",
+        proposed_frontmatter={"description": "release", "tools": []},
+        created_by="test",
+        reason="test",
+    )
+    pipeline.learning_store.update_learning_candidate("candidate-1", draft_skill_name=draft.skill_name, draft_id=draft.draft_id)
+
+    report = asyncio.run(
+        pipeline.evaluate_draft(
+            "candidate-1",
+            draft.skill_name,
+            draft.draft_id,
+            provider_bundle=_bundle(),
+            replay_runner=FakeReplayRunner(
+                baseline_answer="Do not ship. Tests are failing.",
+                candidate_answer="Ship after smoke tests pass.",
+            ),
+        )
+    )
+
+    case = report.case_reports[0]
+    assert case["tool_execution_score"]["baseline_score"] == 0.85
+    assert case["tool_execution_score"]["candidate_score"] == 0.85
+    assert case["baseline_score"] < case["candidate_score"]
+    assert report.tool_mode_summary["score_role"] == "diagnostic_only"
+    assert report.ability_score_summary["score_role"] == "primary"
+    assert report.real_score_avg is not None
+    assert report.synthetic_score_avg is not None
+
+
+def test_synthetic_cases_without_validator_are_not_replay_scored(tmp_path: Path) -> None:
+    pipeline = _pipeline(tmp_path)
+    pipeline.learning_store.update_learning_candidate(
+        "candidate-1",
+        evidence={
+            "eval_cases": [
+                {
+                    "run_id": "synthetic:no-validator",
+                    "task_id": "synthetic-no-validator",
+                    "session_id": "synthetic-eval",
+                    "task_text": "Synthetic task without an oracle.",
+                    "synthetic": True,
+                    "accepted_score": 0.75,
+                }
+            ]
+        },
+    )
+    draft = pipeline.draft_service.create_new_skill_draft(
+        skill_name="release-checklist",
+        proposed_content="# Release\n\nRun tests.",
+        proposed_frontmatter={"description": "release", "tools": []},
+        created_by="test",
+        reason="test",
+    )
+    pipeline.learning_store.update_learning_candidate("candidate-1", draft_skill_name=draft.skill_name, draft_id=draft.draft_id)
+    replay_runner = FakeReplayRunner()
+
+    report = asyncio.run(
+        pipeline.evaluate_draft(
+            "candidate-1",
+            draft.skill_name,
+            draft.draft_id,
+            provider_bundle=_bundle(),
+            replay_runner=replay_runner,
+        )
+    )
+
+    assert "synthetic:no-validator" not in {case["run_id"] for case in report.case_reports}
+    assert all("synthetic:no-validator" not in request.case_id for request in replay_runner.requests)
+    assert report.case_selection_summary["excluded_synthetic_without_validator"] == 1