feat(h5): inject direct-chat history on agent escalation and defer ambiguous routing.

Phase B appends snapshot context when leaving h5direct sessions so Goose no longer starts blind. Phase C lets the 0.72 fallback return null on gated chat text so the existing LLM router can run, flags default off.

Co-authored-by: Cursor <cursoragent@cursor.com>

chat-intent-router.test.mjs

@@ -1054,6 +1054,27 @@ test('classifyWithRules routes memory recall to direct chat on fresh session', (
   assert.equal(lastConversation.route, CHAT_INTENT_ROUTE.DIRECT_CHAT);
 });
 
+test('classifyWithRules defers ambiguous chat text to llm when chat session defer is enabled', () => {
+  const result = classifyWithRules({
+    text: '帮我比较一下英得尔和美的车载冰箱',
+    sessionId: 'h5direct_abc',
+    sessionMessageCount: 3,
+    chatSessionDeferEnabled: true,
+    chatSessionDeferMinTextLength: 12,
+  });
+  assert.equal(result, null);
+});
+
+test('classifyWithRules keeps memory recall on direct chat when defer is enabled', () => {
+  const result = classifyWithRules({
+    text: '你记得我说想去哪儿吗',
+    sessionId: 'h5direct_abc',
+    sessionMessageCount: 3,
+    chatSessionDeferEnabled: true,
+  });
+  assert.equal(result.route, CHAT_INTENT_ROUTE.DIRECT_CHAT);
+});
+
 test('createChatIntentRouter fast-paths news lookup to agent without LLM', async () => {
   let llmCalls = 0;
   const router = createChatIntentRouter({