diff --git a/app/app/ai_register.py b/app/app/ai_register.py
index 85b084d..ca1432d 100644
--- a/app/app/ai_register.py
+++ b/app/app/ai_register.py
@@ -123,10 +123,16 @@ def normalize_register_type(value: str) -> str:
     return register_type
 
 
-def collect_input_lines(text: str | None = None, lines: list[str] | None = None) -> list[str]:
+def collect_input_lines(
+    text: str | None = None,
+    lines: list[str] | None = None,
+    register_type: str = "",
+) -> list[str]:
     if lines is not None:
         return normalize_lines(lines=lines)
     raw = str(text or "")
+    if register_type in {"class_record", "payment"}:
+        return normalize_lines(lines=raw.splitlines())
     if "\n\n" in raw:
         chunks = [item.strip() for item in re.split(r"\n\s*\n", raw) if item.strip()]
         if chunks:
@@ -547,7 +553,7 @@ def preview_register(
     conversation_id: str | None = None,
 ) -> dict[str, Any]:
     normalized_type = normalize_register_type(register_type)
-    source_lines = collect_input_lines(text=text, lines=lines)
+    source_lines = collect_input_lines(text=text, lines=lines, register_type=normalized_type)
     joined_text = "\n\n".join(source_lines)
     conversation_id, session = _session_for(conversation_id, normalized_type, joined_text)
     merged_answers = {**dict(session.get("answers") or {}), **(answers or {})}
diff --git a/app/app/data.py b/app/app/data.py
index cde652d..c010d2d 100644
--- a/app/app/data.py
+++ b/app/app/data.py
@@ -1149,7 +1149,7 @@ def reject_admin_task(tasks_path: Path, task_id: int) -> dict:
     return mark_admin_task(tasks_path, task_id, "rejected")
 
 
-def approve_correction_task(tasks_path: Path, classnotes_path: Path, task_id: int) -> dict:
+def approve_correction_task(tasks_path: Path, classnotes_path: Path, accounts_path: Path, task_id: int) -> dict:
     tasks = read_admin_tasks(tasks_path)
     task = find_admin_task(tasks, task_id)
     if task.get("type") != "class_record_correction":
@@ -1159,13 +1159,25 @@ def approve_correction_task(tasks_path: Path, classnotes_path: Path, task_id: in
 
     original_line = str(task.get("original_line", "")).strip()
     corrected_line = str(task.get("corrected_line", "")).strip()
-    parse_class_record_line(original_line)
+    original = parse_class_record_line(original_line)
     corrected = parse_class_record_line(corrected_line)
     corrected_line = class_record_to_line(corrected)
 
     original_classnotes = classnotes_path.read_text(encoding="utf-8")
     try:
         new_classnotes = replace_class_record_line(original_classnotes, original_line, corrected_line)
+        accounts = read_accounts(accounts_path)
+        updated_accounts = list(accounts)
+        original_account_index = find_account_index(updated_accounts, original.student)
+        updated_accounts[original_account_index] = update_account_remaining(
+            updated_accounts[original_account_index],
+            original.duration_hours,
+        )
+        corrected_account_index = find_account_index(updated_accounts, corrected.student)
+        updated_accounts[corrected_account_index] = update_account_remaining(
+            updated_accounts[corrected_account_index],
+            -corrected.duration_hours,
+        )
     except ValueError as exc:
         task["status"] = "conflict"
         task["updated_at"] = datetime.now().isoformat(timespec="seconds")
@@ -1173,27 +1185,42 @@ def approve_correction_task(tasks_path: Path, classnotes_path: Path, task_id: in
         write_admin_tasks(tasks_path, tasks)
         raise
 
+    original_accounts = accounts_path.read_text(encoding="utf-8")
     original_tasks = json.dumps(tasks, ensure_ascii=False, indent=2) + "\n"
     now = datetime.now().isoformat(timespec="seconds")
     task["status"] = "approved"
     task["updated_at"] = now
     task["reviewed_at"] = now
+    task["original"] = record_to_dict(original)
     task["corrected_line"] = corrected_line
     task["corrected"] = record_to_dict(corrected)
+    changed_account_ids = {
+        updated_accounts[original_account_index].student_id,
+        updated_accounts[corrected_account_index].student_id,
+    }
+    updated_accounts_by_id = {
+        account.student_id: account
+        for account in updated_accounts
+        if account.student_id in changed_account_ids
+    }
+    new_accounts = replace_account_lines(original_accounts, updated_accounts_by_id)
     new_tasks = json.dumps(tasks, ensure_ascii=False, indent=2) + "\n"
 
     backup_dir = create_data_backup(
         "admin-approve-correction",
         {
+            accounts_path: original_accounts,
             classnotes_path: original_classnotes,
             tasks_path: original_tasks,
         },
         [original_line, corrected_line],
     )
     try:
+        atomic_write_text(accounts_path, new_accounts)
         atomic_write_text(classnotes_path, new_classnotes)
         atomic_write_text(tasks_path, new_tasks)
     except Exception:
+        atomic_write_text(accounts_path, original_accounts)
         atomic_write_text(classnotes_path, original_classnotes)
         atomic_write_text(tasks_path, original_tasks)
         raise
@@ -2483,15 +2510,23 @@ def course_summary_heading(summary: dict, existing_headings: set[str]) -> str:
 def save_course_summary_markdown(root: Path, summary: dict) -> dict:
     path = course_summary_path(root, summary)
     existing = path.read_text(encoding="utf-8") if path.exists() else ""
-    existing_headings = set(re.findall(r"^###\s+(.+)$", existing, flags=re.M))
-    existing_compact = re.sub(r"\s+", "", existing)
+    matches = list(COURSE_SUMMARY_HEADING_RE.finditer(existing))
+    existing_headings = {match.group("title").strip() for match in matches}
     body = str(summary.get("body") or "").rstrip()
-    body_compact = re.sub(r"\s+", "", body)
-    if body_compact and body_compact in existing_compact:
-        return {"path": str(path), "added": False}
-
     group_name = str(summary.get("group") or summary["student"])
     heading = course_summary_heading(summary, existing_headings)
+    source_id = str(summary.get("source_id") or "").strip()
+    if source_id and f"来源ID：`{source_id}`" in existing:
+        return {"path": str(path), "added": False}
+    existing_blocks: set[tuple[str, str]] = set()
+    for index, match in enumerate(matches):
+        start = match.end()
+        end = matches[index + 1].start() if index + 1 < len(matches) else len(existing)
+        block_body = existing[start:end].strip()
+        block_body = re.sub(r"^(?:>\s+.*\n)+\s*", "", block_body).strip()
+        existing_blocks.add((match.group("title").strip(), re.sub(r"\s+", "", block_body)))
+    if (heading, re.sub(r"\s+", "", body)) in existing_blocks:
+        return {"path": str(path), "added": False}
     lines: list[str] = []
     if not existing.strip():
         lines.extend([f"# {path.stem}", "", f"## {group_name}", ""])
@@ -2733,7 +2768,50 @@ def approve_course_summary_task(
         raise ValueError("课程小结信息未补齐，不能直接批准入账")
 
     try:
-        result = register_class_record_lines(classnotes_path, accounts_path, line=proposed_line)
+        original_classnotes = classnotes_path.read_text(encoding="utf-8")
+        original_accounts = accounts_path.read_text(encoding="utf-8")
+        original_tasks = json.dumps(tasks, ensure_ascii=False, indent=2) + "\n"
+        record = parse_class_record_line(proposed_line)
+        record_line = class_record_to_line(record)
+        existing_lines = {raw.strip() for raw in original_classnotes.splitlines()}
+        if record_line in existing_lines:
+            raise ValueError(f"上课记录已存在: {record_line}")
+        accounts = read_accounts(accounts_path)
+        updated_accounts = list(accounts)
+        account_index = find_account_index(updated_accounts, record.student)
+        updated_accounts[account_index] = update_account_remaining(
+            updated_accounts[account_index],
+            -record.duration_hours,
+        )
+        separator = "" if not original_classnotes or original_classnotes.endswith("\n") else "\n"
+        new_classnotes = f"{original_classnotes}{separator}{record_line}\n"
+        updated_accounts_by_id = {updated_accounts[account_index].student_id: updated_accounts[account_index]}
+        new_accounts = replace_account_lines(original_accounts, updated_accounts_by_id)
+        task["status"] = "approved"
+        task["updated_at"] = datetime.now().isoformat(timespec="seconds")
+        task["reviewed_at"] = task["updated_at"]
+        task["proposed_line"] = record_line
+        task["registered_line"] = record_line
+        backup_dir = create_data_backup(
+            "admin-approve-course-summary",
+            {
+                accounts_path: original_accounts,
+                classnotes_path: original_classnotes,
+                tasks_path: original_tasks,
+            },
+            [record_line],
+        )
+        task["backup_id"] = backup_dir.name
+        new_tasks = json.dumps(tasks, ensure_ascii=False, indent=2) + "\n"
+        try:
+            atomic_write_text(accounts_path, new_accounts)
+            atomic_write_text(classnotes_path, new_classnotes)
+            atomic_write_text(tasks_path, new_tasks)
+        except Exception:
+            atomic_write_text(accounts_path, original_accounts)
+            atomic_write_text(classnotes_path, original_classnotes)
+            atomic_write_text(tasks_path, original_tasks)
+            raise
     except ValueError as exc:
         task["status"] = "conflict"
         task["updated_at"] = datetime.now().isoformat(timespec="seconds")
@@ -2741,14 +2819,11 @@ def approve_course_summary_task(
         write_admin_tasks(tasks_path, tasks)
         raise
 
-    task["status"] = "approved"
-    task["updated_at"] = datetime.now().isoformat(timespec="seconds")
-    task["reviewed_at"] = task["updated_at"]
-    task["proposed_line"] = proposed_line
-    task["registered_line"] = proposed_line
-    task["backup_id"] = result.get("backup_id", "")
-    write_admin_tasks(tasks_path, tasks)
-    return {"task": task_to_dict(task), "backup_id": result.get("backup_id", "")}
+    try:
+        prune_data_backups(backup_dir.parent)
+    except OSError:
+        pass
+    return {"task": task_to_dict(task), "backup_id": backup_dir.name}
 
 
 def approve_admin_task(
@@ -2759,7 +2834,7 @@ def approve_admin_task(
 ) -> dict:
     task = find_admin_task(read_admin_tasks(tasks_path), task_id)
     if task.get("type") == "class_record_correction":
-        return approve_correction_task(tasks_path, classnotes_path, task_id)
+        return approve_correction_task(tasks_path, classnotes_path, accounts_path, task_id)
     if task.get("type") == "class_record_deletion":
         return approve_deletion_task(tasks_path, classnotes_path, accounts_path, task_id)
     if task.get("type") == "course_summary_review":
diff --git a/app/app/static/admin.js b/app/app/static/admin.js
index 15663f2..f2cd094 100644
--- a/app/app/static/admin.js
+++ b/app/app/static/admin.js
@@ -971,6 +971,20 @@ function clearRegisterPreview(type, resetState = true) {
   if (confirmButton) confirmButton.hidden = true;
 }
 
+function renderStandardLinesPreview(type, lines) {
+  const standardLines = Array.isArray(lines) ? lines : [];
+  if (type === "class_record") {
+    const rows = standardLines
+      .map(
+        (line, index) =>
+          `<li><span>${index + 1}</span><code>${escapeHtml(line)}</code></li>`
+      )
+      .join("");
+    return `<div class="register-preview-head"><strong>标准登记内容</strong><span>共 ${standardLines.length} 条</span></div><ol class="register-line-list">${rows}</ol>`;
+  }
+  return `<strong>标准登记内容</strong><pre>${escapeHtml(standardLines.join("\n\n"))}</pre>`;
+}
+
 function renderRegisterPreview(type, data, statusNode) {
   const previewNode = {
     class_record: classRegisterPreview,
@@ -985,11 +999,12 @@ function renderRegisterPreview(type, data, statusNode) {
   previewNode.hidden = false;
   if (data.status === "ready") {
     const warning = data.warning ? `<p>${escapeHtml(data.warning)}</p>` : "";
-    previewNode.innerHTML = `${warning}<strong>标准登记内容</strong><pre>${escapeHtml((data.standard_lines || []).join("\n\n"))}</pre>`;
+    const standardLines = data.standard_lines || [];
+    previewNode.innerHTML = `${warning}${renderStandardLinesPreview(type, standardLines)}`;
     confirmButton.hidden = false;
     registerPreviewState[type] = {
       conversationId: data.conversation_id,
-      standardLines: data.standard_lines || [],
+      standardLines,
     };
     statusNode.textContent = "已通过脚本生成预览，请确认后写入";
     return;
@@ -1289,5 +1304,4 @@ refreshBtn.addEventListener("click", () => {
 });
 
 loadAdminHealth();
-addSummaryRegisterItem();
 loadAccounts();
diff --git a/app/app/static/styles.css b/app/app/static/styles.css
index 010549d..751918d 100644
--- a/app/app/static/styles.css
+++ b/app/app/static/styles.css
@@ -427,6 +427,62 @@ textarea:focus {
   line-height: 1.45;
 }
 
+.register-preview-head {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 10px;
+}
+
+.register-preview-head span {
+  color: var(--muted);
+  font-size: 12px;
+  font-weight: 700;
+  white-space: nowrap;
+}
+
+.register-line-list {
+  display: grid;
+  gap: 6px;
+  max-height: 220px;
+  margin: 0;
+  padding: 0;
+  overflow: auto;
+  list-style: none;
+}
+
+.register-line-list li {
+  display: grid;
+  grid-template-columns: 28px minmax(0, 1fr);
+  gap: 8px;
+  align-items: start;
+  padding: 8px;
+  border: 1px solid var(--line);
+  border-radius: 6px;
+  background: #fff;
+}
+
+.register-line-list span {
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  min-height: 22px;
+  border-radius: 999px;
+  background: #e9f5f3;
+  color: var(--accent-strong);
+  font-size: 12px;
+  font-weight: 700;
+}
+
+.register-line-list code {
+  color: #344054;
+  font-family: Arial, "Songti SC", SimSun, sans-serif;
+  font-size: 13px;
+  line-height: 1.45;
+  white-space: pre-wrap;
+  overflow-wrap: anywhere;
+}
+
 .register-preview textarea {
   min-height: 72px;
 }