refactor: simplify code across U1-U7 (bug fix + efficiency + reuse + quality)

src/agentkit/chat/skill_routing.py

@@ -98,6 +98,31 @@ def parse_skill_prefix(content: str) -> tuple[str | None, str]:
     return explicit_skill, clean
 
 
+_PROMPT_KEYS = ("identity", "context", "instructions", "constraints", "output_format")
+
+
+def format_preconditions_block(preconditions: list[str], header_level: int = 2) -> str:
+    """格式化激活前置条件文本块。"""
+    header = "#" * header_level
+    lines = [f"{header} Activation Preconditions", "Before executing this skill, verify:"]
+    lines.extend(f"- {p}" for p in preconditions)
+    lines.append(
+        "If any precondition is not met, refuse to execute or ask the user for clarification."
+    )
+    return "\n".join(lines)
+
+
+def collect_prompt_parts(config: Any, with_headers: bool = False) -> list[str]:
+    """从 skill config 的 prompt 字典中收集各部分文本。"""
+    prompt = config.prompt or {}
+    parts: list[str] = []
+    for key in _PROMPT_KEYS:
+        val = prompt.get(key)
+        if val:
+            parts.append(f"### {key.title()}\n{val}" if with_headers else val)
+    return parts
+
+
 def build_skill_system_prompt(skill_config) -> str | None:
     """Build system prompt from skill config's prompt section.
 
@@ -120,12 +145,7 @@ def build_skill_system_prompt(skill_config) -> str | None:
         # 安全守卫：前置条件即使在概要模式下也必须注入
         preconditions = getattr(skill_config, "preconditions", None)
         if preconditions:
-            lines = ["## Activation Preconditions", "Before executing this skill, verify:"]
-            lines.extend(f"- {p}" for p in preconditions)
-            lines.append(
-                "If any precondition is not met, refuse to execute or ask the user for clarification."
-            )
-            return f"{summary}\n\n" + "\n".join(lines)
+            return f"{summary}\n\n" + format_preconditions_block(preconditions)
         # 提示 LLM 可通过 skill_detail 工具加载完整 instructions
         return (
             f"{summary}\n\n"
@@ -135,22 +155,12 @@ def build_skill_system_prompt(skill_config) -> str | None:
     # Level 1+: 全量加载（现有行为）
     if not skill_config.prompt:
         return None
-    prompt_parts = []
-    for key in ("identity", "context", "instructions", "constraints", "output_format"):
-        val = skill_config.prompt.get(key)
-        if val:
-            prompt_parts.append(val)
-    base = "\n\n".join(prompt_parts) if prompt_parts else None
+    base = "\n\n".join(collect_prompt_parts(skill_config)) or None
 
     # v7: 注入激活前置条件（软检查）
     preconditions = getattr(skill_config, "preconditions", None)
     if preconditions:
-        lines = ["## Activation Preconditions", "Before executing this skill, verify:"]
-        lines.extend(f"- {p}" for p in preconditions)
-        lines.append(
-            "If any precondition is not met, refuse to execute or ask the user for clarification."
-        )
-        preconditions_block = "\n".join(lines)
+        preconditions_block = format_preconditions_block(preconditions)
         return f"{base}\n\n{preconditions_block}" if base else preconditions_block
     return base
 

src/agentkit/core/middleware.py

@@ -90,15 +90,12 @@ class MiddlewareChain:
             return await handler(ctx)
 
         # before: 外 → 内
+        # before 异常自然传播，不执行 after 链
         executed_befores: list[Middleware] = []
         current_ctx = ctx
-        try:
         for mw in self._middlewares:
             current_ctx = await mw.before(current_ctx)
             executed_befores.append(mw)
-        except Exception:
-            # before 异常：不执行 after，直接传播
-            raise
 
         # handler
         result = await handler(current_ctx)
@@ -152,12 +149,11 @@ class SummarizationMiddleware:
 class TokenUsageMiddleware:
     """Token 计量中间件 — 记录请求的 token 用量。
 
-    before: 记录起始时间
+    before: 无操作
     after: 从 result 中提取 token usage，记录到 metadata
     """
 
     async def before(self, ctx: RequestContext) -> RequestContext:
-        ctx.metadata["token_usage_start"] = ctx.metadata.get("token_usage_start", 0)
         return ctx
 
     async def after(self, ctx: RequestContext, result: Any) -> Any:
@@ -191,17 +187,18 @@ class LoopDetectionMiddleware:
         if len(trajectory) < self._threshold:
             return result
 
-        # 检查最终 trajectory 中的重复工具调用模式
+        # 检查最终 trajectory 中的重复工具调用模式（只取尾部窗口）
+        recent = trajectory[-self._window_size :] if trajectory else []
         tool_calls = [
             (step.get("tool_name", ""), step.get("arguments_hash", ""))
-            for step in trajectory
+            for step in recent
             if isinstance(step, dict) and "tool_name" in step
         ]
         if not tool_calls:
             return result
 
         # 滑动窗口检测
-        window = tool_calls[-self._window_size :]
+        window = tool_calls
         unique = set(window)
         if len(unique) < len(window) and len(window) - len(unique) >= self._threshold - 1:
             logger.warning(

src/agentkit/core/react.py

@@ -98,8 +98,7 @@ async def _ensure_async_iterable(obj: Any, label: str = "<obj>"):
                 yield item
             return
         raise TypeError(
-            f"{label}: awaited value is not async iterable "
-            f"(got {type(resolved).__name__})"
+            f"{label}: awaited value is not async iterable (got {type(resolved).__name__})"
         )
 
     # Case 3: anything else — surface a clear, actionable error rather
@@ -218,19 +217,17 @@ class ReActEngine:
 
         ponytail: 精确 hash 匹配，不做语义相似度。
         """
+        hash_to_name: dict[str, str] = {}
         for tc in tool_calls:
             args_str = json.dumps(tc.arguments, sort_keys=True, default=str)
-            h = hash(f"{tc.name}:{args_str}")
-            self._loop_window.append(str(h))
+            h = str(hash(f"{tc.name}:{args_str}"))
+            self._loop_window.append(h)
+            hash_to_name[h] = tc.name
 
         counts = Counter(self._loop_window)
         for h, count in counts.items():
             if count >= self._loop_threshold:
-                # Find the tool name for this hash
-                for tc in tool_calls:
-                    args_str = json.dumps(tc.arguments, sort_keys=True, default=str)
-                    if str(hash(f"{tc.name}:{args_str}")) == h:
-                        return tc.name
+                return hash_to_name.get(h)
         return None
 
     async def execute(
@@ -1162,6 +1159,13 @@ class ReActEngine:
                             )
 
                     # 记录 LLM 调用步骤
+                    if trace_recorder is not None:
+                        trace_recorder.record_step(
+                            step=step,
+                            action="llm_call",
+                            duration_ms=llm_duration_ms,
+                            tokens_used=step_tokens,
+                        )
 
                     # Record assistant message
                     assistant_msg: dict[str, Any] = {

src/agentkit/orchestrator/checkpoint.py

@@ -116,13 +116,9 @@ class PipelineCheckpoint:
         # 尝试写入 Redis
         if self._redis is not None:
             try:
-                await self._redis.set(
-                    self._plan_key(plan_id), json.dumps(plan_dict), ex=self._ttl
-                )
+                await self._redis.set(self._plan_key(plan_id), json.dumps(plan_dict), ex=self._ttl)
             except Exception as e:
-                logger.warning(
-                    f"PipelineCheckpoint.save_plan Redis failed for plan {plan_id}: {e}"
-                )
+                logger.warning(f"PipelineCheckpoint.save_plan Redis failed for plan {plan_id}: {e}")
 
     async def load_plan(self, plan_id: str) -> dict[str, Any] | None:
         """加载完整 plan JSON。"""
@@ -133,9 +129,7 @@ class PipelineCheckpoint:
                 if raw:
                     return json.loads(raw)
             except Exception as e:
-                logger.warning(
-                    f"PipelineCheckpoint.load_plan Redis failed for plan {plan_id}: {e}"
-                )
+                logger.warning(f"PipelineCheckpoint.load_plan Redis failed for plan {plan_id}: {e}")
         # 内存降级（检查 TTL 过期）
         entry = self._memory_plans.get(plan_id)
         if entry is None:
@@ -217,14 +211,17 @@ class PipelineCheckpoint:
                 if not phase_ids:
                     # Redis 无数据，检查内存（过滤过期）
                     return [
-                        c
-                        for c in self._memory.get(plan_id, [])
-                        if not self._is_expired(c.saved_at)
+                        c for c in self._memory.get(plan_id, []) if not self._is_expired(c.saved_at)
                     ]
 
-                results: list[CheckpointData] = []
+                # 批量 GET（pipeline 避免 N+1 往返）
+                pipe = self._redis.pipeline()
                 for phase_id in phase_ids:
-                    raw = await self._redis.get(self._key(plan_id, phase_id))
+                    pipe.get(self._key(plan_id, phase_id))
+                raws = await pipe.execute()
+
+                results: list[CheckpointData] = []
+                for raw in raws:
                     if raw:
                         data = json.loads(raw)
                         results.append(CheckpointData.from_dict(data))
@@ -255,6 +252,4 @@ class PipelineCheckpoint:
                 pipe.delete(self._plan_key(plan_id))
                 await pipe.execute()
             except Exception as e:
-                logger.warning(
-                    f"PipelineCheckpoint.clear Redis failed for plan {plan_id}: {e}"
-                )
+                logger.warning(f"PipelineCheckpoint.clear Redis failed for plan {plan_id}: {e}")

src/agentkit/skills/skill_detail.py

@@ -14,6 +14,8 @@ from __future__ import annotations
 import logging
 from typing import Any
 
+from agentkit.chat.skill_routing import collect_prompt_parts, format_preconditions_block
+from agentkit.core.exceptions import SkillNotFoundError
 from agentkit.tools.base import Tool
 
 logger = logging.getLogger(__name__)
@@ -91,7 +93,7 @@ class SkillDetailTool(Tool):
         try:
             skill = self._registry.get(query)
             return self._format_skill_full(skill)
-        except Exception:
+        except SkillNotFoundError:
             pass  # 非精确匹配，降级到关键词搜索
 
         # 关键词搜索：匹配 skill 名称和描述
@@ -118,21 +120,12 @@ class SkillDetailTool(Tool):
     def _format_skill_full(skill: Any) -> dict[str, Any]:
         """格式化 skill 的完整 instructions 供 LLM 使用。"""
         config = skill.config
-        prompt_parts: list[str] = []
-        for key in ("identity", "context", "instructions", "constraints", "output_format"):
-            val = (config.prompt or {}).get(key)
-            if val:
-                prompt_parts.append(f"### {key.title()}\n{val}")
+        prompt_parts = collect_prompt_parts(config, with_headers=True)
 
         # v7: 注入激活前置条件（安全守卫，即使在渐进加载模式下也不应省略）
         preconditions = getattr(config, "preconditions", None)
         if preconditions:
-            lines = ["### Activation Preconditions", "Before executing this skill, verify:"]
-            lines.extend(f"- {p}" for p in preconditions)
-            lines.append(
-                "If any precondition is not met, refuse to execute or ask the user for clarification."
-            )
-            prompt_parts.append("\n".join(lines))
+            prompt_parts.append(format_preconditions_block(preconditions, header_level=3))
 
         return {
             "name": config.name,

tests/unit/test_middleware.py

@@ -428,12 +428,12 @@ class TestTokenUsageMiddleware:
     """TokenUsageMiddleware 测试。"""
 
     @pytest.mark.asyncio
-    async def test_before_initializes_metadata(self) -> None:
-        """before 初始化 token_usage_start。"""
+    async def test_before_returns_ctx_unchanged(self) -> None:
+        """before 返回 ctx 不做修改（token 计量在 after 中完成）。"""
         mw = TokenUsageMiddleware()
         ctx = _make_ctx()
-        await mw.before(ctx)
-        assert "token_usage_start" in ctx.metadata
+        result = await mw.before(ctx)
+        assert result is ctx
 
     @pytest.mark.asyncio
     async def test_after_extracts_usage_from_result(self) -> None: