fix: 提升 CLAUDE.md 指令权重 — 独立 project-instructions + deferred tools 位置调整

- prependUserContext: 将 claudeMd 从通用 <system-reminder> 提取为独立的
  <project-instructions> 用户消息，不带免责声明，置于消息列表最前面
- queryModel: deferred tools 消息从 prepend 改为 append，避免抢占
  project-instructions 的最高权重位置；标签规范化为 <system-reminder>

Co-Authored-By: glm-5-turbo <zai-org@claude-code-best.win>

src/services/api/claude.ts

@@ -1407,12 +1407,14 @@ async function* queryModel(
       .sort()
       .join('\n')
     if (deferredToolList) {
+      // Append to the end of the messages array (not prepend) so it
+      // never抢占 <project-instructions> (CLAUDE.md) at the front.
       messagesForAPI = [
+        ...messagesForAPI,
         createUserMessage({
-          content: `<available-deferred-tools>\n${deferredToolList}\n</available-deferred-tools>\nTo invoke any tool listed above, use ExecuteExtraTool with {"tool_name": "<name>", "params": {...}}. This is the ONLY way to call deferred tools — do not read source code or analyze implementation, just call ExecuteExtraTool directly.`,
+          content: `<system-reminder>\n<available-deferred-tools>\n${deferredToolList}\n</available-deferred-tools>\nTo invoke any tool listed above, use ExecuteExtraTool with {"tool_name": "<name>", "params": {...}}. This is the ONLY way to call deferred tools — do not read source code or analyze implementation, just call ExecuteExtraTool directly.\n</system-reminder>`,
           isMeta: true,
         }),
-        ...messagesForAPI,
       ]
     }
   }

src/utils/api.ts

@@ -452,19 +452,36 @@ export function prependUserContext(
     return messages
   }
 
-  return [
-    createUserMessage({
-      content: `<system-reminder>\nAs you answer the user's questions, you can use the following context:\n${Object.entries(
-        context,
-      )
-        .map(([key, value]) => `# ${key}\n${value}`)
-        .join('\n')}
+  // Extract claudeMd as a dedicated high-weight user message so it isn't
+  // buried inside the generic <system-reminder> with the "may or may not be
+  // relevant" disclaimer, which would degrade its instructional weight.
+  const { claudeMd, ...rest } = context
+  const result: Message[] = []
+
+  if (claudeMd) {
+    result.push(
+      createUserMessage({
+        content: `<project-instructions>\n${claudeMd}\n</project-instructions>\n`,
+        isMeta: true,
+      }),
+    )
+  }
+
+  const restEntries = Object.entries(rest)
+  if (restEntries.length > 0) {
+    result.push(
+      createUserMessage({
+        content: `<system-reminder>\nAs you answer the user's questions, you can use the following context:\n${restEntries
+          .map(([key, value]) => `# ${key}\n${value}`)
+          .join('\n')}
 
       IMPORTANT: this context may or may not be relevant to your tasks. You should not respond to this context unless it is highly relevant to your task.\n</system-reminder>\n`,
-      isMeta: true,
-    }),
-    ...messages,
-  ]
+        isMeta: true,
+      }),
+    )
+  }
+
+  return [...result, ...messages]
 }
 
 /**