fix: 应用代码审查三项发现

- frontend: .tab 补 position:relative,修复激活 tab 下划线定位缺失定位上下文
- core/retry.py: 删除全项目零调用的重试模块(避免"看似生效实际未接入"的误判)
- llm/validation.py: KNOWN_AGENT_NAMES 改为从 orchestrator.SPECIALIST_SPECS 派生,单一权威源

src/llm/validation.py

@@ -10,8 +10,10 @@ from pydantic import BaseModel, Field, field_validator, model_validator
 
 logger = logging.getLogger(__name__)
 
-# 已知的 5 个专家 agent 名(与 orchestrator.SPECIALIST_SPECS 保持一致)
-KNOWN_AGENT_NAMES: tuple[str, ...] = ("form", "stats", "home_away", "injuries", "h2h")
+# 单一权威源:从 orchestrator.SPECIALIST_SPECS 派生,避免两端独立定义导致静默偏离
+from src.llm.agents.orchestrator import SPECIALIST_SPECS
+
+KNOWN_AGENT_NAMES: tuple[str, ...] = tuple(spec.name for spec in SPECIALIST_SPECS)
 
 
 class AgentReportSchema(BaseModel):