From 12c7a4bdd9f151c948587d085752597232897382 Mon Sep 17 00:00:00 2001
From: JOJO <1498581755@qq.com>
Date: Tue, 17 Mar 2026 22:43:51 +0800
Subject: [PATCH] fix: remove legacy file edit tags

---
 CLAUDE.md                                     |  19 +-
 README.md                                     |   2 +-
 core/main_terminal_parts/context.py           |   3 +
 core/tool_config.py                           |   2 -
 core/web_terminal.py                          |   7 -
 modules/file_manager.py                       |   2 +-
 modules/personalization_manager.py            |   7 +
 scripts/api_tool_role_experiment.py           |  25 +-
 server/app_legacy.py                          |   8 +-
 server/chat_flow_helpers.py                   |   8 +-
 server/chat_flow_pending_writes.py            | 546 ------------------
 server/chat_flow_stream_loop.py               | 248 +-------
 server/chat_flow_task_main.py                 | 242 +-------
 server/state.py                               |   2 +-
 static/src/app.ts                             |   2 -
 static/src/app/methods/history.ts             |  46 +-
 .../components/chat/actions/ToolAction.vue    |  65 ---
 .../components/chat/actions/toolRenderers.ts  |  77 ---
 .../personalization/PersonalizationDrawer.vue |  25 +-
 static/src/stores/chat.ts                     |  20 -
 static/src/stores/monitor.ts                  |   2 -
 static/src/stores/personalization.ts          |   3 +
 static/src/utils/chatDisplay.ts               |   6 -
 static/src/utils/icons.ts                     |   2 -
 utils/api_client.py                           |  32 -
 25 files changed, 107 insertions(+), 1294 deletions(-)
 delete mode 100644 server/chat_flow_pending_writes.py

diff --git a/CLAUDE.md b/CLAUDE.md
index 1675512..7aa28c6 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -96,24 +96,9 @@ tail -f logs/container_stats.log
 
 ## Important Implementation Details
 
-### File Operations with Streaming
+### File Operations
 
-**append_to_file** 和 **modify_file** 使用特殊的流式输出格式：
-
-```
-<<<APPEND:path/to/file>>>
-文件内容...
-<<<END_APPEND>>>
-
-<<<MODIFY:path/to/file>>>
-[replace:1]
-<<OLD>>原文内容<<END>>
-<<NEW>>新内容<<END>>
-[/replace]
-<<<END_MODIFY>>>
-```
-
-处理逻辑在 `web_server.py` 的 `handle_task_with_sender` 中，通过检测标记并在流式输出中即时执行。
+文件写入统一使用 `write_file`/`edit_file` 工具完成：`write_file` 负责覆盖或追加内容，`edit_file` 负责精确字符串替换，不再依赖历史版本的流式标签标记。
 
 ### Container Architecture
 
diff --git a/README.md b/README.md
index 1c0bbdb..d9e2421 100644
--- a/README.md
+++ b/README.md
@@ -113,7 +113,7 @@ npm run dev
 | --- | --- |
 | `/help` | CLI 指令列表 |
 | `/status` | 查看系统/容器资源状态 |
-| `read_file` / `modify_file` | 读写项目文件（自动通过容器代理） |
+| `read_file` / `write_file` / `edit_file` | 读写项目文件（自动通过容器代理） |
 | `terminal_session` / `terminal_input` | 管理多终端会话 |
 | `run_command` / `run_python` | 工具容器快速执行命令/Python 代码 |
 | `todo_*`, `update_memory` | 维护待办与长期记忆 |
diff --git a/core/main_terminal_parts/context.py b/core/main_terminal_parts/context.py
index fa3e8fa..d4ed3dc 100644
--- a/core/main_terminal_parts/context.py
+++ b/core/main_terminal_parts/context.py
@@ -258,6 +258,9 @@ class MainTerminalContextMixin:
                         self.context_manager._build_content_with_images(conv["content"], images, videos)
                         if (images or videos) else conv["content"]
                     )
+                    # 调试：记录所有 system 消息
+                    if conv["role"] == "system":
+                        logger.info(f"[DEBUG build_messages] 添加 system 消息: content前50字={conv['content'][:50]}")
                     messages.append({
                         "role": conv["role"],
                         "content": content_payload
diff --git a/core/tool_config.py b/core/tool_config.py
index e507d62..a8f2a46 100644
--- a/core/tool_config.py
+++ b/core/tool_config.py
@@ -33,8 +33,6 @@ TOOL_CATEGORIES: Dict[str, ToolCategory] = {
             "create_file",
             "write_file",
             "edit_file",
-            "append_to_file",
-            "modify_file",
             "delete_file",
             "rename_file",
             "create_folder",
diff --git a/core/web_terminal.py b/core/web_terminal.py
index 76daf3a..af5d22f 100644
--- a/core/web_terminal.py
+++ b/core/web_terminal.py
@@ -395,13 +395,6 @@ class WebTerminal(MainTerminal):
                 'status': 'reading',
                 'detail': f'读取文件({read_type}): {arguments.get("path", "未知路径")}'
             })
-        elif tool_name == "modify_file":
-            path = arguments.get("path", "未知路径")
-            self.broadcast('tool_status', {
-                'tool': tool_name,
-                'status': 'modifying',
-                'detail': f'准备修改文件: {path}'
-            })
         elif tool_name == "delete_file":
             self.broadcast('tool_status', {
                 'tool': tool_name,
diff --git a/modules/file_manager.py b/modules/file_manager.py
index d0bab80..c7c8a72 100644
--- a/modules/file_manager.py
+++ b/modules/file_manager.py
@@ -1076,7 +1076,7 @@ class FileManager:
             if old_text is None or new_text is None:
                 block_result["status"] = "error"
                 block_result["reason"] = "缺少 OLD 或 NEW 内容"
-                block_result["hint"] = "请确保粘贴的补丁包含成对的 <<<OLD>>> / <<<NEW>>> 标记。"
+                block_result["hint"] = "请确保补丁包含成对的 OLD/NEW 段落。"
                 failed_details.append({"index": index, "reason": "缺少 OLD/NEW 标记"})
                 results.append(block_result)
                 continue
diff --git a/modules/personalization_manager.py b/modules/personalization_manager.py
index 25bd697..14a3340 100644
--- a/modules/personalization_manager.py
+++ b/modules/personalization_manager.py
@@ -39,6 +39,7 @@ DEFAULT_PERSONALIZATION_CONFIG: Dict[str, Any] = {
     "default_run_mode": None,
     "auto_generate_title": True,
     "tool_intent_enabled": True,
+    "skill_hints_enabled": False,  # Skill 提示系统开关（默认关闭）
     "default_model": "kimi-k2.5",
     "image_compression": "original",  # original / 1080p / 720p / 540p
     "silent_tool_disable": False,  # 禁用工具时不向模型插入提示
@@ -145,6 +146,12 @@ def sanitize_personalization_payload(
     else:
         base["tool_intent_enabled"] = bool(base.get("tool_intent_enabled"))
 
+    # Skill 提示系统开关
+    if "skill_hints_enabled" in data:
+        base["skill_hints_enabled"] = bool(data.get("skill_hints_enabled"))
+    else:
+        base["skill_hints_enabled"] = bool(base.get("skill_hints_enabled"))
+
     if "disabled_tool_categories" in data:
         base["disabled_tool_categories"] = _sanitize_tool_categories(data.get("disabled_tool_categories"), allowed_tool_categories)
     else:
diff --git a/scripts/api_tool_role_experiment.py b/scripts/api_tool_role_experiment.py
index 87a1cc8..ca018fc 100644
--- a/scripts/api_tool_role_experiment.py
+++ b/scripts/api_tool_role_experiment.py
@@ -48,40 +48,41 @@ def load_conversation_messages(path: Path) -> List[Dict[str, Any]]:
 
 
 def minimal_tool_definitions() -> List[Dict[str, Any]]:
-    """返回涵盖 append/modify 的最小工具定义集合。"""
+    """返回涵盖 write/edit 的最小工具定义集合。"""
     return [
         {
             "type": "function",
             "function": {
-                "name": "append_to_file",
+                "name": "write_file",
                 "description": (
-                    "准备向文件追加大段内容。调用后系统会发放 <<<APPEND:path>>>…<<<END_APPEND>>> "
-                    "格式的写入窗口，AI 必须在窗口内一次性输出需要追加的全部内容。"
+                    "将内容写入本地文件系统；append 为 True 时追加到末尾，False 时覆盖原文件。"
                 ),
                 "parameters": {
                     "type": "object",
                     "properties": {
-                        "path": {"type": "string", "description": "目标文件的相对路径"},
-                        "reason": {"type": "string", "description": "为什么需要追加（可选）"}
+                        "file_path": {"type": "string", "description": "目标文件的相对路径"},
+                        "content": {"type": "string", "description": "要写入的内容"},
+                        "append": {"type": "boolean", "description": "是否追加到末尾", "default": False}
                     },
-                    "required": ["path"]
+                    "required": ["file_path", "content"]
                 }
             }
         },
         {
             "type": "function",
             "function": {
-                "name": "modify_file",
+                "name": "edit_file",
                 "description": (
-                    "准备替换文件中的指定内容。模型必须在 <<<MODIFY:path>>>…<<<END_MODIFY>>> "
-                    "结构内输出若干 [replace:n] 补丁块，每块包含 <<OLD>> 原文 和 <<NEW>> 新内容。"
+                    "在文件中执行精确字符串替换，要求 old_string 与文件内容精确匹配。"
                 ),
                 "parameters": {
                     "type": "object",
                     "properties": {
-                        "path": {"type": "string", "description": "目标文件的相对路径"}
+                        "file_path": {"type": "string", "description": "目标文件的相对路径"},
+                        "old_string": {"type": "string", "description": "需要替换的原文"},
+                        "new_string": {"type": "string", "description": "替换后的新内容"}
                     },
-                    "required": ["path"]
+                    "required": ["file_path", "old_string", "new_string"]
                 }
             }
         }
diff --git a/server/app_legacy.py b/server/app_legacy.py
index b3fdffa..01c340d 100644
--- a/server/app_legacy.py
+++ b/server/app_legacy.py
@@ -1173,16 +1173,10 @@ def apply_thinking_schedule(terminal: WebTerminal):
         client.skip_thinking_next_call = False
         return
     state = get_thinking_state(terminal)
-    awaiting_writes = getattr(terminal, "pending_append_request", None) or getattr(terminal, "pending_modify_request", None)
-    if awaiting_writes:
-        client.skip_thinking_next_call = True
-        state["suppress_next"] = False
-        debug_log("[Thinking] 检测到写入窗口请求，跳过思考。")
-        return
     if state.get("suppress_next"):
         client.skip_thinking_next_call = True
         state["suppress_next"] = False
-        debug_log("[Thinking] 由于写入窗口，下一次跳过思考。")
+        debug_log("[Thinking] 已标记跳过思考模式。")
         return
     if state.get("force_next"):
         client.force_thinking_next_call = True
diff --git a/server/chat_flow_helpers.py b/server/chat_flow_helpers.py
index 9f8b621..815c124 100644
--- a/server/chat_flow_helpers.py
+++ b/server/chat_flow_helpers.py
@@ -137,16 +137,10 @@ def apply_thinking_schedule(terminal: WebTerminal, default_interval: int, debug_
         return
 
     state = get_thinking_state(terminal)
-    awaiting_writes = getattr(terminal, "pending_append_request", None) or getattr(terminal, "pending_modify_request", None)
-    if awaiting_writes:
-        client.skip_thinking_next_call = True
-        state["suppress_next"] = False
-        debug_logger("[Thinking] 检测到写入窗口请求，跳过思考。")
-        return
     if state.get("suppress_next"):
         client.skip_thinking_next_call = True
         state["suppress_next"] = False
-        debug_logger("[Thinking] 由于写入窗口，下一次跳过思考。")
+        debug_logger("[Thinking] 已标记跳过思考模式。")
         return
     if state.get("force_next"):
         client.force_thinking_next_call = True
diff --git a/server/chat_flow_pending_writes.py b/server/chat_flow_pending_writes.py
deleted file mode 100644
index c131f83..0000000
--- a/server/chat_flow_pending_writes.py
+++ /dev/null
@@ -1,546 +0,0 @@
-from __future__ import annotations
-
-import json
-import re
-from typing import Dict, List, Optional
-
-
-async def finalize_pending_append(*, pending_append, append_probe_buffer: str, response_text: str, stream_completed: bool, finish_reason: str = None, web_terminal, sender, debug_log):
-    """在流式输出结束后处理追加写入"""
-
-    result = {
-        "handled": False,
-        "success": False,
-        "summary": None,
-        "summary_message": None,
-        "tool_content": None,
-        "tool_call_id": None,
-        "path": None,
-        "forced": False,
-        "error": None,
-        "assistant_content": response_text,
-        "lines": 0,
-        "bytes": 0,
-        "finish_reason": finish_reason,
-        "appended_content": "",
-        "assistant_metadata": None
-    }
-
-    if not pending_append:
-        return result, pending_append, append_probe_buffer
-
-    state = pending_append
-    path = state.get("path")
-    tool_call_id = state.get("tool_call_id")
-    buffer = state.get("buffer", "")
-    start_marker = state.get("start_marker")
-    end_marker = state.get("end_marker")
-    start_idx = state.get("content_start")
-    end_idx = state.get("end_index")
-
-    display_id = state.get("display_id")
-
-    result.update({
-        "handled": True,
-        "path": path,
-        "tool_call_id": tool_call_id,
-        "display_id": display_id
-    })
-
-    if path is None or tool_call_id is None:
-        error_msg = "append_to_file 状态不完整，缺少路径或ID。"
-        debug_log(error_msg)
-        result["error"] = error_msg
-        result["summary_message"] = error_msg
-        result["tool_content"] = json.dumps({
-            "success": False,
-            "error": error_msg
-        }, ensure_ascii=False)
-        if display_id:
-            sender('update_action', {
-                'id': display_id,
-                'status': 'failed',
-                'preparing_id': tool_call_id,
-                'message': error_msg
-            })
-        pending_append = None
-        return result, pending_append, append_probe_buffer
-
-    if start_idx is None:
-        error_msg = f"未检测到格式正确的开始标识 {start_marker}。"
-        debug_log(error_msg)
-        result["error"] = error_msg
-        result["summary_message"] = error_msg
-        result["tool_content"] = json.dumps({
-            "success": False,
-            "path": path,
-            "error": error_msg
-        }, ensure_ascii=False)
-        if display_id:
-            sender('update_action', {
-                'id': display_id,
-                'status': 'failed',
-                'preparing_id': tool_call_id,
-                'message': error_msg
-            })
-        pending_append = None
-        return result, pending_append, append_probe_buffer
-
-    forced = False
-    if end_idx is None:
-        forced = True
-        # 查找下一个<<<，否则使用整个缓冲结尾
-        remaining = buffer[start_idx:]
-        next_marker = remaining.find("<<<", len(end_marker))
-        if next_marker != -1:
-            end_idx = start_idx + next_marker
-        else:
-            end_idx = len(buffer)
-
-    content = buffer[start_idx:end_idx]
-    if content.startswith('\n'):
-        content = content[1:]
-
-    if not content:
-        error_msg = "未检测到需要追加的内容，请严格按照<<<APPEND:path>>>...<<<END_APPEND>>>格式输出。"
-        debug_log(error_msg)
-        result["error"] = error_msg
-        result["forced"] = forced
-        result["tool_content"] = json.dumps({
-            "success": False,
-            "path": path,
-            "error": error_msg
-        }, ensure_ascii=False)
-        if display_id:
-            sender('update_action', {
-                'id': display_id,
-                'status': 'failed',
-                'preparing_id': tool_call_id,
-                'message': error_msg
-            })
-        pending_append = None
-        return result, pending_append, append_probe_buffer
-
-    assistant_message_lines = []
-    if start_marker:
-        assistant_message_lines.append(start_marker)
-    assistant_message_lines.append(content)
-    if not forced and end_marker:
-        assistant_message_lines.append(end_marker)
-    assistant_message_text = "\n".join(assistant_message_lines)
-    result["assistant_content"] = assistant_message_text
-    assistant_metadata = {
-        "append_payload": {
-            "path": path,
-            "tool_call_id": tool_call_id,
-            "forced": forced,
-            "has_end_marker": not forced
-        }
-    }
-    result["assistant_metadata"] = assistant_metadata
-
-    write_result = web_terminal.file_manager.append_file(path, content)
-    if write_result.get("success"):
-        bytes_written = len(content.encode('utf-8'))
-        line_count = content.count('\n')
-        if content and not content.endswith('\n'):
-            line_count += 1
-
-        summary = f"已向 {path} 追加 {line_count} 行（{bytes_written} 字节）"
-        if forced:
-            summary += "。未检测到 <<<END_APPEND>>> 标记，系统已在流结束处完成写入。如内容未完成，请重新调用 append_to_file 并按标准格式补充；如已完成，可继续后续步骤。"
-
-        result.update({
-            "success": True,
-            "summary": summary,
-            "summary_message": summary,
-            "forced": forced,
-            "lines": line_count,
-            "bytes": bytes_written,
-            "appended_content": content,
-            "tool_content": json.dumps({
-                "success": True,
-                "path": path,
-                "lines": line_count,
-                "bytes": bytes_written,
-                "forced": forced,
-                "message": summary,
-                "finish_reason": finish_reason
-            }, ensure_ascii=False)
-        })
-
-        assistant_meta_payload = result["assistant_metadata"]["append_payload"]
-        assistant_meta_payload["lines"] = line_count
-        assistant_meta_payload["bytes"] = bytes_written
-        assistant_meta_payload["success"] = True
-
-        summary_payload = {
-            "success": True,
-            "path": path,
-            "lines": line_count,
-            "bytes": bytes_written,
-            "forced": forced,
-            "message": summary
-        }
-
-        if display_id:
-            sender('update_action', {
-                'id': display_id,
-                'status': 'completed',
-                'result': summary_payload,
-                'preparing_id': tool_call_id,
-                'message': summary
-            })
-
-        debug_log(f"追加写入完成: {summary}")
-    else:
-        error_msg = write_result.get("error", "追加写入失败")
-        result.update({
-            "error": error_msg,
-            "summary_message": error_msg,
-            "forced": forced,
-            "appended_content": content,
-            "tool_content": json.dumps({
-                "success": False,
-                "path": path,
-                "error": error_msg,
-                "finish_reason": finish_reason
-            }, ensure_ascii=False)
-        })
-        debug_log(f"追加写入失败: {error_msg}")
-
-        if result["assistant_metadata"]:
-            assistant_meta_payload = result["assistant_metadata"]["append_payload"]
-            assistant_meta_payload["lines"] = content.count('\n') + (0 if content.endswith('\n') or not content else 1)
-            assistant_meta_payload["bytes"] = len(content.encode('utf-8'))
-            assistant_meta_payload["success"] = False
-
-        failure_payload = {
-            "success": False,
-            "path": path,
-            "error": error_msg,
-            "forced": forced
-        }
-
-        if display_id:
-            sender('update_action', {
-                'id': display_id,
-                'status': 'completed',
-                'result': failure_payload,
-                'preparing_id': tool_call_id,
-                'message': error_msg
-            })
-
-    pending_append = None
-    append_probe_buffer = ""
-    if hasattr(web_terminal, "pending_append_request"):
-        web_terminal.pending_append_request = None
-    return result, pending_append, append_probe_buffer
-
-
-
-async def finalize_pending_modify(*, pending_modify, modify_probe_buffer: str, response_text: str, stream_completed: bool, finish_reason: str = None, web_terminal, sender, debug_log):
-    """在流式输出结束后处理修改写入"""
-
-    result = {
-        "handled": False,
-        "success": False,
-        "path": None,
-        "tool_call_id": None,
-        "display_id": None,
-        "total_blocks": 0,
-        "completed_blocks": [],
-        "failed_blocks": [],
-        "forced": False,
-        "details": [],
-        "error": None,
-        "assistant_content": response_text,
-        "assistant_metadata": None,
-        "tool_content": None,
-        "summary_message": None,
-        "finish_reason": finish_reason
-    }
-
-    if not pending_modify:
-        return result, pending_modify, modify_probe_buffer
-
-    state = pending_modify
-    path = state.get("path")
-    tool_call_id = state.get("tool_call_id")
-    display_id = state.get("display_id")
-    start_marker = state.get("start_marker")
-    end_marker = state.get("end_marker")
-    buffer = state.get("buffer", "")
-    raw_buffer = state.get("raw_buffer", "")
-    end_index = state.get("end_index")
-
-    result.update({
-        "handled": True,
-        "path": path,
-        "tool_call_id": tool_call_id,
-        "display_id": display_id
-    })
-
-    if not state.get("start_seen"):
-        error_msg = "未检测到格式正确的 <<<MODIFY:path>>> 标记。"
-        debug_log(error_msg)
-        result["error"] = error_msg
-        result["summary_message"] = error_msg
-        result["tool_content"] = json.dumps({
-            "success": False,
-            "path": path,
-            "error": error_msg,
-            "finish_reason": finish_reason
-        }, ensure_ascii=False)
-        if display_id:
-            sender('update_action', {
-                'id': display_id,
-                'status': 'failed',
-                'preparing_id': tool_call_id,
-                'message': error_msg
-            })
-        if hasattr(web_terminal, "pending_modify_request"):
-            web_terminal.pending_modify_request = None
-        pending_modify = None
-        modify_probe_buffer = ""
-        return result, pending_modify, modify_probe_buffer
-
-    forced = end_index is None
-    apply_text = buffer if forced else buffer[:end_index]
-    raw_content = raw_buffer if forced else raw_buffer[:len(start_marker) + end_index + len(end_marker)]
-    if raw_content:
-        result["assistant_content"] = raw_content
-
-    blocks_info = []
-    block_reports = {}
-    detected_indices = set()
-    block_pattern = re.compile(r"\[replace:(\d+)\](.*?)\[/replace\]", re.DOTALL)
-    structure_warnings: List[str] = []
-    structure_detail_entries: List[Dict] = []
-
-    def record_structure_warning(message: str, hint: Optional[str] = None):
-        """记录结构性缺陷，便于给出更具体的反馈。"""
-        if message in structure_warnings:
-            return
-        structure_warnings.append(message)
-        structure_detail_entries.append({
-            "index": 0,
-            "status": "failed",
-            "reason": message,
-            "removed_lines": 0,
-            "added_lines": 0,
-            "hint": hint or "请严格按照模板输出：[replace:n] + <<OLD>>/<<NEW>> + [/replace]，并使用 <<<END_MODIFY>>> 收尾。"
-        })
-
-    def extract_segment(body: str, tag: str):
-        marker = f"<<{tag}>>"
-        end_tag = "<<END>>"
-        start_pos = body.find(marker)
-        if start_pos == -1:
-            return None, f"缺少 {marker}"
-        start_pos += len(marker)
-        if body[start_pos:start_pos+2] == "\r\n":
-            start_pos += 2
-        elif body[start_pos:start_pos+1] == "\n":
-            start_pos += 1
-        end_pos = body.find(end_tag, start_pos)
-        if end_pos == -1:
-            return None, f"缺少 {end_tag}"
-        segment = body[start_pos:end_pos]
-        return segment, None
-
-    for match in block_pattern.finditer(apply_text):
-        try:
-            index = int(match.group(1))
-        except ValueError:
-            continue
-        body = match.group(2)
-        if index in detected_indices:
-            continue
-        detected_indices.add(index)
-        block_reports[index] = {
-            "index": index,
-            "status": "pending",
-            "reason": None,
-            "removed_lines": 0,
-            "added_lines": 0,
-            "hint": None
-        }
-        old_content, old_error = extract_segment(body, "OLD")
-        new_content, new_error = extract_segment(body, "NEW")
-        if old_error or new_error:
-            reason = old_error or new_error
-            block_reports[index]["status"] = "failed"
-            block_reports[index]["reason"] = reason
-        blocks_info.append({
-            "index": index,
-            "old": old_content,
-            "new": new_content,
-            "error": old_error or new_error
-        })
-
-    if not blocks_info:
-        has_replace_start = bool(re.search(r"\[replace:\s*\d+\]", apply_text))
-        has_replace_end = "[/replace]" in apply_text
-        has_old_tag = "<<OLD>>" in apply_text
-        has_new_tag = "<<NEW>>" in apply_text
-
-        if has_replace_start and not has_replace_end:
-            record_structure_warning("检测到 [replace:n] 标记但缺少对应的 [/replace] 结束标记。")
-        if has_replace_end and not has_replace_start:
-            record_structure_warning("检测到 [/replace] 结束标记但缺少对应的 [replace:n] 起始标记。")
-
-        old_tags = len(re.findall(r"<<OLD>>", apply_text))
-        completed_old_tags = len(re.findall(r"<<OLD>>[\s\S]*?<<END>>", apply_text))
-        if old_tags and completed_old_tags < old_tags:
-            record_structure_warning("检测到 <<OLD>> 段落但未看到对应的 <<END>> 结束标记。")
-
-        new_tags = len(re.findall(r"<<NEW>>", apply_text))
-        completed_new_tags = len(re.findall(r"<<NEW>>[\s\S]*?<<END>>", apply_text))
-        if new_tags and completed_new_tags < new_tags:
-            record_structure_warning("检测到 <<NEW>> 段落但未看到对应的 <<END>> 结束标记。")
-
-        if (has_replace_start or has_replace_end or has_old_tag or has_new_tag) and not structure_warnings:
-            record_structure_warning("检测到部分补丁标记，但整体结构不完整，请严格按照模板填写所有标记。")
-
-    total_blocks = len(blocks_info)
-    result["total_blocks"] = total_blocks
-    if forced:
-        debug_log("未检测到 <<<END_MODIFY>>>，将在流结束处执行已识别的修改块。")
-        result["forced"] = True
-
-    blocks_to_apply = [
-        {"index": block["index"], "old": block["old"], "new": block["new"]}
-        for block in blocks_info
-        if block["error"] is None and block["old"] is not None and block["new"] is not None
-    ]
-
-    # 记录格式残缺的块
-    for block in blocks_info:
-        if block["error"]:
-            idx = block["index"]
-            block_reports[idx]["status"] = "failed"
-            block_reports[idx]["reason"] = block["error"]
-            block_reports[idx]["hint"] = "请检查补丁块的 OLD/NEW 标记是否完整，必要时复用 terminal_snapshot 或终端命令重新调整。"
-
-    apply_result = {}
-    if blocks_to_apply:
-        apply_result = web_terminal.file_manager.apply_modify_blocks(path, blocks_to_apply)
-    else:
-        apply_result = {"success": False, "completed": [], "failed": [], "results": [], "write_performed": False, "error": None}
-
-    block_result_map = {item["index"]: item for item in apply_result.get("results", [])}
-
-    for block in blocks_info:
-        idx = block["index"]
-        report = block_reports.get(idx)
-        if report is None:
-            continue
-        if report["status"] == "failed":
-            continue
-        block_apply = block_result_map.get(idx)
-        if not block_apply:
-            report["status"] = "failed"
-            report["reason"] = "未执行，可能未找到匹配原文"
-            report["hint"] = report.get("hint") or "请确认 OLD 文本与文件内容完全一致；若多次失败，可改用终端命令/Python 进行精准替换。"
-            continue
-        status = block_apply.get("status")
-        report["removed_lines"] = block_apply.get("removed_lines", 0)
-        report["added_lines"] = block_apply.get("added_lines", 0)
-        if block_apply.get("hint"):
-            report["hint"] = block_apply.get("hint")
-        if status == "success":
-            report["status"] = "completed"
-        elif status == "not_found":
-            report["status"] = "failed"
-            report["reason"] = block_apply.get("reason") or "未找到匹配的原文"
-            if not report.get("hint"):
-                report["hint"] = "请使用 terminal_snapshot/grep -n 校验原文，或在说明后改用 run_command/python 精确替换。"
-        else:
-            report["status"] = "failed"
-            report["reason"] = block_apply.get("reason") or "替换失败"
-            if not report.get("hint"):
-                report["hint"] = block_apply.get("hint") or "若多次尝试仍失败，可考虑利用终端命令或 Python 小脚本完成此次修改。"
-
-    completed_blocks = sorted([idx for idx, rep in block_reports.items() if rep["status"] == "completed"])
-    failed_blocks = sorted([idx for idx, rep in block_reports.items() if rep["status"] != "completed"])
-
-    result["completed_blocks"] = completed_blocks
-    result["failed_blocks"] = failed_blocks
-    details = sorted(block_reports.values(), key=lambda x: x["index"])
-    if structure_detail_entries:
-        details = structure_detail_entries + details
-    result["details"] = details
-
-    summary_parts = []
-    if total_blocks == 0:
-        summary_parts.append("未检测到有效的修改块，未执行任何修改。")
-        summary_parts.extend(structure_warnings)
-    else:
-        if not completed_blocks and failed_blocks:
-            summary_parts.append(f"共检测到 {total_blocks} 个修改块，全部未执行。")
-        elif completed_blocks and not failed_blocks:
-            summary_parts.append(f"共 {total_blocks} 个修改块全部完成。")
-        else:
-            summary_parts.append(
-                f"共检测到 {total_blocks} 个修改块，其中成功 {len(completed_blocks)} 个，失败 {len(failed_blocks)} 个。"
-            )
-    if forced:
-        summary_parts.append("未检测到 <<<END_MODIFY>>> 标记，系统已在流结束处执行补丁。")
-    if apply_result.get("error"):
-        summary_parts.append(apply_result["error"])
-
-    matching_note = "提示：补丁匹配基于完整文本，包含注释和空白符，请确保 <<<OLD>>> 段落与文件内容逐字一致。如果修改成功，请忽略，如果失败，请明确原文后再次尝试。"
-    summary_parts.append(matching_note)
-    summary_message = " ".join(summary_parts).strip()
-    result["summary_message"] = summary_message
-    result["success"] = bool(completed_blocks) and not failed_blocks and apply_result.get("error") is None
-
-    tool_payload = {
-        "success": result["success"],
-        "path": path,
-        "total_blocks": total_blocks,
-        "completed": completed_blocks,
-        "failed": [
-            {
-                "index": rep["index"],
-                "reason": rep.get("reason"),
-                "hint": rep.get("hint")
-            }
-            for rep in result["details"] if rep["status"] != "completed"
-        ],
-        "forced": forced,
-        "message": summary_message,
-        "finish_reason": finish_reason,
-        "details": result["details"]
-    }
-    if apply_result.get("error"):
-        tool_payload["error"] = apply_result["error"]
-
-    result["tool_content"] = json.dumps(tool_payload, ensure_ascii=False)
-    result["assistant_metadata"] = {
-        "modify_payload": {
-            "path": path,
-            "total_blocks": total_blocks,
-            "completed": completed_blocks,
-            "failed": failed_blocks,
-            "forced": forced,
-            "details": result["details"]
-        }
-    }
-
-    if display_id:
-        sender('update_action', {
-            'id': display_id,
-            'status': 'completed' if result["success"] else 'failed',
-            'result': tool_payload,
-            'preparing_id': tool_call_id,
-            'message': summary_message
-        })
-
-    pending_modify = None
-    modify_probe_buffer = ""
-    if hasattr(web_terminal, "pending_modify_request"):
-        web_terminal.pending_modify_request = None
-    return result, pending_modify, modify_probe_buffer
-
diff --git a/server/chat_flow_stream_loop.py b/server/chat_flow_stream_loop.py
index c6b1af4..c3ee766 100644
--- a/server/chat_flow_stream_loop.py
+++ b/server/chat_flow_stream_loop.py
@@ -3,19 +3,17 @@ from __future__ import annotations
 import asyncio
 import json
 import time
-import re
 from typing import Any, Dict, Optional
 
 from config.model_profiles import get_model_profile
 
 from .utils_common import debug_log, brief_log, log_backend_chunk
 from .chat_flow_runner_helpers import extract_intent_from_partial
-from .chat_flow_pending_writes import finalize_pending_append, finalize_pending_modify
 from .chat_flow_task_support import wait_retry_delay, cancel_pending_tools
 from .state import get_stop_flag, clear_stop_flag
 
 
-async def run_streaming_attempts(*, web_terminal, messages, tools, sender, client_sid: str, username: str, conversation_id: Optional[str], current_iteration: int, max_api_retries: int, retry_delay_seconds: int, pending_append, append_probe_buffer: str, pending_modify, modify_probe_buffer: str, detected_tool_intent: Dict[str, str], full_response: str, tool_calls: list, current_thinking: str, detected_tools: Dict[str, str], last_usage_payload, in_thinking: bool, thinking_started: bool, thinking_ended: bool, text_started: bool, text_has_content: bool, text_streaming: bool, text_chunk_index: int, last_text_chunk_time, chunk_count: int, reasoning_chunks: int, content_chunks: int, tool_chunks: int, append_result: Dict[str, Any], modify_result: Dict[str, Any], last_finish_reason: Optional[str], accumulated_response: str) -> Dict[str, Any]:
+async def run_streaming_attempts(*, web_terminal, messages, tools, sender, client_sid: str, username: str, conversation_id: Optional[str], current_iteration: int, max_api_retries: int, retry_delay_seconds: int, detected_tool_intent: Dict[str, str], full_response: str, tool_calls: list, current_thinking: str, detected_tools: Dict[str, str], last_usage_payload, in_thinking: bool, thinking_started: bool, thinking_ended: bool, text_started: bool, text_has_content: bool, text_streaming: bool, text_chunk_index: int, last_text_chunk_time, chunk_count: int, reasoning_chunks: int, content_chunks: int, tool_chunks: int, last_finish_reason: Optional[str], accumulated_response: str) -> Dict[str, Any]:
     api_error = None
     for api_attempt in range(max_api_retries + 1):
         api_error = None
@@ -37,13 +35,8 @@ async def run_streaming_attempts(*, web_terminal, messages, tools, sender, clien
             reasoning_chunks = 0
             content_chunks = 0
             tool_chunks = 0
-            append_result = {"handled": False}
-            modify_result = {"handled": False}
             last_finish_reason = None
 
-        append_break_triggered = False
-        modify_break_triggered = False
-
         # 收集流式响应
         async for chunk in web_terminal.api_client.chat(messages, tools, stream=True):
             chunk_count += 1
@@ -54,10 +47,6 @@ async def run_streaming_attempts(*, web_terminal, messages, tools, sender, clien
                 stop_requested = client_stop_info.get('stop', False) if isinstance(client_stop_info, dict) else client_stop_info
                 if stop_requested:
                     debug_log(f"检测到停止请求，中断流处理")
-                    if pending_append:
-                        append_result, pending_append, append_probe_buffer = await finalize_pending_append(pending_append=pending_append, append_probe_buffer=append_probe_buffer, response_text=full_response, stream_completed=False, finish_reason="user_stop", web_terminal=web_terminal, sender=sender, debug_log=debug_log)
-                    if pending_modify:
-                        modify_result, pending_modify, modify_probe_buffer = await finalize_pending_modify(pending_modify=pending_modify, modify_probe_buffer=modify_probe_buffer, response_text=full_response, stream_completed=False, finish_reason="user_stop", web_terminal=web_terminal, sender=sender, debug_log=debug_log)
                     cancel_pending_tools(tool_calls_list=tool_calls, sender=sender, messages=messages)
                     sender('task_stopped', {
                         'message': '命令执行被用户取消',
@@ -83,13 +72,7 @@ async def run_streaming_attempts(*, web_terminal, messages, tools, sender, clien
                         "reasoning_chunks": reasoning_chunks,
                         "content_chunks": content_chunks,
                         "tool_chunks": tool_chunks,
-                        "append_result": append_result,
-                        "modify_result": modify_result,
                         "last_finish_reason": last_finish_reason,
-                        "pending_append": pending_append,
-                        "append_probe_buffer": append_probe_buffer,
-                        "pending_modify": pending_modify,
-                        "modify_probe_buffer": modify_probe_buffer,
                         "accumulated_response": accumulated_response,
                     }
 
@@ -158,168 +141,6 @@ async def run_streaming_attempts(*, web_terminal, messages, tools, sender, clien
                         sender('thinking_end', {'full_content': current_thinking})
                         await asyncio.sleep(0.1)
 
-
-                    expecting_modify = bool(pending_modify) or bool(getattr(web_terminal, "pending_modify_request", None))
-                    expecting_append = bool(pending_append) or bool(getattr(web_terminal, "pending_append_request", None))
-
-                    if pending_modify:
-                        if not pending_modify.get("start_seen"):
-                            probe_buffer = pending_modify.get("probe_buffer", "") + content
-                            if len(probe_buffer) > 10000:
-                                probe_buffer = probe_buffer[-10000:]
-                            marker = pending_modify.get("start_marker")
-                            marker_index = probe_buffer.find(marker)
-                            if marker_index == -1:
-                                pending_modify["probe_buffer"] = probe_buffer
-                                continue
-                            after_marker = marker_index + len(marker)
-                            remainder = probe_buffer[after_marker:]
-                            pending_modify["buffer"] = remainder
-                            pending_modify["raw_buffer"] = marker + remainder
-                            pending_modify["start_seen"] = True
-                            pending_modify["detected_blocks"] = set()
-                            pending_modify["probe_buffer"] = ""
-                            if pending_modify.get("display_id"):
-                                sender('update_action', {
-                                    'id': pending_modify["display_id"],
-                                    'status': 'running',
-                                    'preparing_id': pending_modify.get("tool_call_id"),
-                                    'message': f"正在修改 {pending_modify['path']}..."
-                                })
-                        else:
-                            pending_modify["buffer"] += content
-                            pending_modify["raw_buffer"] += content
-
-                        if pending_modify.get("start_seen"):
-                            block_text = pending_modify["buffer"]
-                            for match in re.finditer(r"\[replace:(\d+)\]", block_text):
-                                try:
-                                    block_index = int(match.group(1))
-                                except ValueError:
-                                    continue
-                                detected_blocks = pending_modify.setdefault("detected_blocks", set())
-                                if block_index not in detected_blocks:
-                                    detected_blocks.add(block_index)
-                                    if pending_modify.get("display_id"):
-                                        sender('update_action', {
-                                            'id': pending_modify["display_id"],
-                                            'status': 'running',
-                                            'preparing_id': pending_modify.get("tool_call_id"),
-                                            'message': f"正在对 {pending_modify['path']} 进行第 {block_index} 处修改..."
-                                        })
-
-                        if pending_modify.get("start_seen"):
-                            end_pos = pending_modify["buffer"].find(pending_modify["end_marker"])
-                            if end_pos != -1:
-                                pending_modify["end_index"] = end_pos
-                                modify_break_triggered = True
-                                debug_log("检测到<<<END_MODIFY>>>，即将终止流式输出并应用修改")
-                                break
-                        continue
-                    elif expecting_modify:
-                        modify_probe_buffer += content
-                        if len(modify_probe_buffer) > 10000:
-                            modify_probe_buffer = modify_probe_buffer[-10000:]
-
-                        marker_match = re.search(r"<<<MODIFY:\s*([\s\S]*?)>>>", modify_probe_buffer)
-                        if marker_match:
-                            detected_raw_path = marker_match.group(1)
-                            detected_path = detected_raw_path.strip()
-                            marker_full = marker_match.group(0)
-                            after_marker_index = modify_probe_buffer.find(marker_full) + len(marker_full)
-                            remainder = modify_probe_buffer[after_marker_index:]
-                            modify_probe_buffer = ""
-
-                            if not detected_path:
-                                debug_log("检测到 MODIFY 起始标记但路径为空，忽略。")
-                                continue
-
-                            pending_modify = {
-                                "path": detected_path,
-                                "tool_call_id": None,
-                                "buffer": remainder,
-                                "raw_buffer": marker_full + remainder,
-                                "start_marker": marker_full,
-                                "end_marker": "<<<END_MODIFY>>>",
-                                "start_seen": True,
-                                "end_index": None,
-                                "display_id": None,
-                                "detected_blocks": set()
-                            }
-                            if hasattr(web_terminal, "pending_modify_request"):
-                                web_terminal.pending_modify_request = {"path": detected_path}
-                            debug_log(f"直接检测到modify起始标记，构建修改缓冲: {detected_path}")
-
-                            end_pos = pending_modify["buffer"].find(pending_modify["end_marker"])
-                            if end_pos != -1:
-                                pending_modify["end_index"] = end_pos
-                                modify_break_triggered = True
-                                debug_log("检测到<<<END_MODIFY>>>，即将终止流式输出并应用修改")
-                                break
-                        continue
-
-                    if pending_append:
-                        pending_append["buffer"] += content
-
-                        if pending_append.get("content_start") is None:
-                            marker_index = pending_append["buffer"].find(pending_append["start_marker"])
-                            if marker_index != -1:
-                                pending_append["content_start"] = marker_index + len(pending_append["start_marker"])
-                                debug_log(f"检测到追加起始标识: {pending_append['start_marker']}")
-
-                        if pending_append.get("content_start") is not None:
-                            end_index = pending_append["buffer"].find(
-                                pending_append["end_marker"],
-                                pending_append["content_start"]
-                            )
-                            if end_index != -1:
-                                pending_append["end_index"] = end_index
-                                append_break_triggered = True
-                                debug_log("检测到<<<END_APPEND>>>，即将终止流式输出并写入文件")
-                                break
-
-                        # 继续累积追加内容
-                        continue
-                    elif expecting_append:
-                        append_probe_buffer += content
-                        # 限制缓冲区大小防止过长
-                        if len(append_probe_buffer) > 10000:
-                            append_probe_buffer = append_probe_buffer[-10000:]
-
-                        marker_match = re.search(r"<<<APPEND:\s*([\s\S]*?)>>>", append_probe_buffer)
-                        if marker_match:
-                            detected_raw_path = marker_match.group(1)
-                            detected_path = detected_raw_path.strip()
-                            if not detected_path:
-                                append_probe_buffer = append_probe_buffer[marker_match.end():]
-                                continue
-                            marker_full = marker_match.group(0)
-                            after_marker_index = append_probe_buffer.find(marker_full) + len(marker_full)
-                            remainder = append_probe_buffer[after_marker_index:]
-                            append_probe_buffer = ""
-                            pending_append = {
-                                "path": detected_path,
-                                "tool_call_id": None,
-                                "buffer": remainder,
-                                "start_marker": marker_full,
-                                "end_marker": "<<<END_APPEND>>>",
-                                "content_start": 0,
-                                "end_index": None,
-                                "display_id": None
-                            }
-                            if hasattr(web_terminal, "pending_append_request"):
-                                web_terminal.pending_append_request = {"path": detected_path}
-                            debug_log(f"直接检测到append起始标记，构建追加缓冲: {detected_path}")
-                            # 检查是否立即包含结束标记
-                            if pending_append["buffer"]:
-                                end_index = pending_append["buffer"].find(pending_append["end_marker"], pending_append["content_start"])
-                                if end_index != -1:
-                                    pending_append["end_index"] = end_index
-                                    append_break_triggered = True
-                                    debug_log("检测到<<<END_APPEND>>>，即将终止流式输出并写入文件")
-                                    break
-                            continue
-
                     if not text_started:
                         text_started = True
                         text_streaming = True
@@ -327,27 +148,26 @@ async def run_streaming_attempts(*, web_terminal, messages, tools, sender, clien
                         brief_log("模型输出了内容")
                         await asyncio.sleep(0.05)
 
-                    if not pending_append:
-                        full_response += content
-                        accumulated_response += content
-                        text_has_content = True
-                        emit_time = time.time()
-                        elapsed = 0.0 if last_text_chunk_time is None else emit_time - last_text_chunk_time
-                        last_text_chunk_time = emit_time
-                        text_chunk_index += 1
-                        log_backend_chunk(
-                            conversation_id,
-                            current_iteration,
-                            text_chunk_index,
-                            elapsed,
-                            len(content),
-                            content[:32]
-                        )
-                        sender('text_chunk', {
-                            'content': content,
-                            'index': text_chunk_index,
-                            'elapsed': elapsed
-                        })
+                    full_response += content
+                    accumulated_response += content
+                    text_has_content = True
+                    emit_time = time.time()
+                    elapsed = 0.0 if last_text_chunk_time is None else emit_time - last_text_chunk_time
+                    last_text_chunk_time = emit_time
+                    text_chunk_index += 1
+                    log_backend_chunk(
+                        conversation_id,
+                        current_iteration,
+                        text_chunk_index,
+                        elapsed,
+                        len(content),
+                        content[:32]
+                    )
+                    sender('text_chunk', {
+                        'content': content,
+                        'index': text_chunk_index,
+                        'elapsed': elapsed
+                    })
 
             # 收集工具调用 - 实时发送准备状态
             if "tool_calls" in delta:
@@ -472,13 +292,7 @@ async def run_streaming_attempts(*, web_terminal, messages, tools, sender, clien
                     "reasoning_chunks": reasoning_chunks,
                     "content_chunks": content_chunks,
                     "tool_chunks": tool_chunks,
-                    "append_result": append_result,
-                    "modify_result": modify_result,
                     "last_finish_reason": last_finish_reason,
-                    "pending_append": pending_append,
-                    "append_probe_buffer": append_probe_buffer,
-                    "pending_modify": pending_modify,
-                    "modify_probe_buffer": modify_probe_buffer,
                     "accumulated_response": accumulated_response,
                 }
 
@@ -546,8 +360,6 @@ async def run_streaming_attempts(*, web_terminal, messages, tools, sender, clien
                 and not full_response
                 and not tool_calls
                 and not current_thinking
-                and not pending_append
-                and not pending_modify
             )
             sender('error', {
                 'message': error_message,
@@ -590,13 +402,7 @@ async def run_streaming_attempts(*, web_terminal, messages, tools, sender, clien
                         "reasoning_chunks": reasoning_chunks,
                         "content_chunks": content_chunks,
                         "tool_chunks": tool_chunks,
-                        "append_result": append_result,
-                        "modify_result": modify_result,
                         "last_finish_reason": last_finish_reason,
-                        "pending_append": pending_append,
-                        "append_probe_buffer": append_probe_buffer,
-                        "pending_modify": pending_modify,
-                        "modify_probe_buffer": modify_probe_buffer,
                         "accumulated_response": accumulated_response,
                     }
                 continue
@@ -620,13 +426,7 @@ async def run_streaming_attempts(*, web_terminal, messages, tools, sender, clien
                 "reasoning_chunks": reasoning_chunks,
                 "content_chunks": content_chunks,
                 "tool_chunks": tool_chunks,
-                "append_result": append_result,
-                "modify_result": modify_result,
                 "last_finish_reason": last_finish_reason,
-                "pending_append": pending_append,
-                "append_probe_buffer": append_probe_buffer,
-                "pending_modify": pending_modify,
-                "modify_probe_buffer": modify_probe_buffer,
                 "accumulated_response": accumulated_response,
             }
         break
@@ -650,12 +450,6 @@ async def run_streaming_attempts(*, web_terminal, messages, tools, sender, clien
         "reasoning_chunks": reasoning_chunks,
         "content_chunks": content_chunks,
         "tool_chunks": tool_chunks,
-        "append_result": append_result,
-        "modify_result": modify_result,
         "last_finish_reason": last_finish_reason,
-        "pending_append": pending_append,
-        "append_probe_buffer": append_probe_buffer,
-        "pending_modify": pending_modify,
-        "modify_probe_buffer": modify_probe_buffer,
         "accumulated_response": accumulated_response,
     }
diff --git a/server/chat_flow_task_main.py b/server/chat_flow_task_main.py
index 45878f2..95550db 100644
--- a/server/chat_flow_task_main.py
+++ b/server/chat_flow_task_main.py
@@ -39,6 +39,7 @@ from modules.personalization_manager import (
     THINKING_INTERVAL_MIN,
     THINKING_INTERVAL_MAX,
 )
+from modules.skill_hint_manager import SkillHintManager
 from modules.upload_security import UploadSecurityError
 from modules.user_manager import UserWorkspace
 from modules.usage_tracker import QUOTA_DEFAULTS
@@ -62,7 +63,7 @@ from .utils_common import (
     CHUNK_FRONTEND_LOG_FILE,
     STREAMING_DEBUG_LOG_FILE,
 )
-from .security import rate_limited, format_tool_result_notice, compact_web_search_result, consume_socket_token, prune_socket_tokens, validate_csrf_request, requires_csrf_protection, get_csrf_token
+from .security import rate_limited, compact_web_search_result, consume_socket_token, prune_socket_tokens, validate_csrf_request, requires_csrf_protection, get_csrf_token
 from .monitor import cache_monitor_snapshot, get_cached_monitor_snapshot
 from .extensions import socketio
 from .state import (
@@ -125,7 +126,6 @@ from .chat_flow_runtime import (
     detect_malformed_tool_call,
 )
 
-from .chat_flow_pending_writes import finalize_pending_append, finalize_pending_modify
 from .chat_flow_task_support import process_sub_agent_updates
 from .chat_flow_tool_loop import execute_tool_calls
 from .chat_flow_stream_loop import run_streaming_attempts
@@ -465,7 +465,30 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
     history_len_before = len(getattr(web_terminal.context_manager, "conversation_history", []) or [])
     is_first_user_message = history_len_before == 0
     web_terminal.context_manager.add_conversation("user", message, images=images, videos=videos)
-    
+
+    # Skill 提示系统：检测关键词并在用户消息之后插入 system 消息
+    try:
+        personal_config = load_personalization_config(workspace.data_dir)
+        skill_hints_enabled = personal_config.get("skill_hints_enabled", False)
+
+        if skill_hints_enabled and message:
+            hint_manager = SkillHintManager()
+            hint_manager.set_enabled(True)
+            hint_messages = hint_manager.build_hint_messages(message)
+
+            # 将提示消息插入到对话历史中（在用户消息之后）
+            for hint_msg in hint_messages:
+                debug_log(f"[Skill Hints] 插入提示消息: {hint_msg['content'][:100]}")
+                web_terminal.context_manager.add_conversation(
+                    "system",
+                    hint_msg["content"]
+                )
+                # 验证插入后的消息
+                last_msg = web_terminal.context_manager.conversation_history[-1]
+                debug_log(f"[Skill Hints] 插入后验证 - role: {last_msg.get('role')}, content: {last_msg.get('content')[:100]}")
+    except Exception as exc:
+        debug_log(f"Skill hints 处理失败: {exc}")
+
     if is_first_user_message and getattr(web_terminal, "context_manager", None):
         try:
             personal_config = load_personalization_config(workspace.data_dir)
@@ -551,10 +574,6 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
     max_api_retries = 4
     retry_delay_seconds = 10
     
-    pending_append = None  # {"path": str, "tool_call_id": str, "buffer": str, ...}
-    append_probe_buffer = ""
-    pending_modify = None  # {"path": str, "tool_call_id": str, "buffer": str, ...}
-    modify_probe_buffer = ""
 
     iteration = 0
     while max_iterations is None or iteration < max_iterations:
@@ -596,8 +615,6 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
         reasoning_chunks = 0
         content_chunks = 0
         tool_chunks = 0
-        append_result = {"handled": False}
-        modify_result = {"handled": False}
         last_finish_reason = None
 
         thinking_expected = web_terminal.api_client.get_current_thinking_mode()
@@ -638,10 +655,6 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
             current_iteration=current_iteration,
             max_api_retries=max_api_retries,
             retry_delay_seconds=retry_delay_seconds,
-            pending_append=pending_append,
-            append_probe_buffer=append_probe_buffer,
-            pending_modify=pending_modify,
-            modify_probe_buffer=modify_probe_buffer,
             detected_tool_intent=detected_tool_intent,
             full_response=full_response,
             tool_calls=tool_calls,
@@ -660,8 +673,6 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
             reasoning_chunks=reasoning_chunks,
             content_chunks=content_chunks,
             tool_chunks=tool_chunks,
-            append_result=append_result,
-            modify_result=modify_result,
             last_finish_reason=last_finish_reason,
             accumulated_response=accumulated_response,
         )
@@ -685,13 +696,7 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
         reasoning_chunks = stream_result["reasoning_chunks"]
         content_chunks = stream_result["content_chunks"]
         tool_chunks = stream_result["tool_chunks"]
-        append_result = stream_result["append_result"]
-        modify_result = stream_result["modify_result"]
         last_finish_reason = stream_result["last_finish_reason"]
-        pending_append = stream_result["pending_append"]
-        append_probe_buffer = stream_result["append_probe_buffer"]
-        pending_modify = stream_result["pending_modify"]
-        modify_probe_buffer = stream_result["modify_probe_buffer"]
         accumulated_response = stream_result["accumulated_response"]
 
         # 流结束后的处理
@@ -704,11 +709,6 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
         debug_log(f"  收集到的正文: {len(full_response)} 字符")
         debug_log(f"  收集到的工具: {len(tool_calls)} 个")
         
-        if not append_result["handled"] and pending_append:
-            append_result, pending_append, append_probe_buffer = await finalize_pending_append(pending_append=pending_append, append_probe_buffer=append_probe_buffer, response_text=full_response, stream_completed=True, finish_reason=last_finish_reason, web_terminal=web_terminal, sender=sender, debug_log=debug_log)
-        if not modify_result["handled"] and pending_modify:
-            modify_result, pending_modify, modify_probe_buffer = await finalize_pending_modify(pending_modify=pending_modify, modify_probe_buffer=modify_probe_buffer, response_text=full_response, stream_completed=True, finish_reason=last_finish_reason, web_terminal=web_terminal, sender=sender, debug_log=debug_log)
-        
         # 结束未完成的流
         if in_thinking and not thinking_ended:
             sender('thinking_end', {'full_content': current_thinking})
@@ -716,7 +716,7 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
             
         
         # 确保text_end事件被发送
-        if text_started and text_has_content and not append_result["handled"] and not modify_result["handled"]:
+        if text_started and text_has_content:
             debug_log(f"发送text_end事件，完整内容长度: {len(full_response)}")
             sender('text_end', {'full_content': full_response})
             await asyncio.sleep(0.1)
@@ -725,159 +725,6 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
             if full_response.strip():
                 debug_log(f"流式文本内容长度: {len(full_response)} 字符")
         
-        if append_result["handled"]:
-            append_metadata = append_result.get("assistant_metadata")
-            append_content_text = append_result.get("assistant_content")
-            if append_content_text:
-                web_terminal.context_manager.add_conversation(
-                    "assistant",
-                    append_content_text,
-                    metadata=append_metadata
-                )
-                debug_log("💾 增量保存：追加正文快照")
-                
-                payload_info = append_metadata.get("append_payload") if append_metadata else {}
-                sender('append_payload', {
-                    'path': payload_info.get("path") or append_result.get("path"),
-                    'forced': payload_info.get("forced", False),
-                    'lines': payload_info.get("lines"),
-                    'bytes': payload_info.get("bytes"),
-                    'tool_call_id': payload_info.get("tool_call_id") or append_result.get("tool_call_id"),
-                    'success': payload_info.get("success", append_result.get("success", False)),
-                    'conversation_id': conversation_id
-                })
-            
-            if append_result["tool_content"]:
-                tool_call_id = append_result.get("tool_call_id") or f"append_{int(time.time() * 1000)}"
-                system_notice = format_tool_result_notice("append_to_file", tool_call_id, append_result["tool_content"])
-                web_terminal.context_manager.add_conversation("system", system_notice)
-                append_result["tool_call_id"] = tool_call_id
-                debug_log("💾 增量保存：append_to_file 工具结果（system 通知）")
-            
-            finish_reason = append_result.get("finish_reason")
-            path_for_prompt = append_result.get("path")
-            need_follow_prompt = (
-                finish_reason == "length" or
-                append_result.get("forced") or
-                not append_result.get("success")
-            )
-
-            if need_follow_prompt and path_for_prompt:
-                prompt_lines = [
-                    f"append_to_file 在处理 {path_for_prompt} 时未完成，需要重新发起写入。"
-                ]
-                if finish_reason == "length":
-                    prompt_lines.append(
-                        "上一次输出达到系统单次输出上限，已写入的内容已保存。"
-                    )
-                if append_result.get("forced"):
-                    prompt_lines.append(
-                        "收到的内容缺少 <<<END_APPEND>>> 标记，系统依据流式结束位置落盘。"
-                    )
-                if not append_result.get("success"):
-                    prompt_lines.append("系统未能识别有效的追加标记。")
-                prompt_lines.append(
-                    "请再次调用 append_to_file 工具获取新的写入窗口，并在工具调用的输出中遵循以下格式："
-                )
-                prompt_lines.append(f"<<<APPEND:{path_for_prompt}>>>")
-                prompt_lines.append("...填写剩余正文，如内容已完成可留空...")
-                prompt_lines.append("<<<END_APPEND>>>")
-                prompt_lines.append("不要在普通回复中粘贴上述标记，必须通过 append_to_file 工具发送。")
-                follow_prompt = "\n".join(prompt_lines)
-                messages.append({
-                    "role": "system",
-                    "content": follow_prompt
-                })
-                web_terminal.context_manager.add_conversation("system", follow_prompt)
-                debug_log("已注入追加任务提示")
-
-        if append_result["handled"] and append_result.get("forced") and append_result.get("success"):
-            mark_force_thinking(web_terminal, reason="append_forced_finish")
-        if append_result["handled"] and not append_result.get("success"):
-            sender('system_message', {
-                'content': f'⚠️ 追加写入失败：{append_result.get("error")}'
-            })
-            maybe_mark_failure_from_message(web_terminal, f'⚠️ 追加写入失败：{append_result.get("error")}')
-            mark_force_thinking(web_terminal, reason="append_failed")
-        
-        if modify_result["handled"]:
-            modify_metadata = modify_result.get("assistant_metadata")
-            modify_content_text = modify_result.get("assistant_content")
-            if modify_content_text:
-                web_terminal.context_manager.add_conversation(
-                    "assistant",
-                    modify_content_text,
-                    metadata=modify_metadata
-                )
-                debug_log("💾 增量保存：修改正文快照")
-                
-                payload_info = modify_metadata.get("modify_payload") if modify_metadata else {}
-                sender('modify_payload', {
-                    'path': payload_info.get("path") or modify_result.get("path"),
-                    'total': payload_info.get("total_blocks") or modify_result.get("total_blocks"),
-                    'completed': payload_info.get("completed") or modify_result.get("completed_blocks"),
-                    'failed': payload_info.get("failed") or modify_result.get("failed_blocks"),
-                    'forced': payload_info.get("forced", modify_result.get("forced", False)),
-                    'success': modify_result.get("success", False),
-                    'conversation_id': conversation_id
-                })
-            
-            if modify_result["tool_content"]:
-                tool_call_id = modify_result.get("tool_call_id") or f"modify_{int(time.time() * 1000)}"
-                system_notice = format_tool_result_notice("modify_file", tool_call_id, modify_result["tool_content"])
-                web_terminal.context_manager.add_conversation("system", system_notice)
-                modify_result["tool_call_id"] = tool_call_id
-                debug_log("💾 增量保存：modify_file 工具结果（system 通知）")
-            
-            path_for_prompt = modify_result.get("path")
-            failed_blocks = modify_result.get("failed_blocks") or []
-            need_follow_prompt = modify_result.get("forced") or bool(failed_blocks)
-            
-            if need_follow_prompt and path_for_prompt:
-                prompt_lines = [
-                    f"modify_file 在处理 {path_for_prompt} 时未完成，需要重新发起补丁。"
-                ]
-                if modify_result.get("forced"):
-                    prompt_lines.append(
-                        "刚才的内容缺少 <<<END_MODIFY>>> 标记，系统仅应用了已识别的部分。"
-                    )
-                if failed_blocks:
-                    failed_text = "、".join(str(idx) for idx in failed_blocks)
-                    prompt_lines.append(f"以下补丁未成功：第 {failed_text} 处。")
-                prompt_lines.append(
-                    "请再次调用 modify_file 工具，并在新的工具调用中按以下模板提供完整补丁："
-                )
-                prompt_lines.append(f"<<<MODIFY:{path_for_prompt}>>>")
-                prompt_lines.append("[replace:序号]")
-                prompt_lines.append("<<OLD>>")
-                prompt_lines.append("...原文（必须逐字匹配，包含全部缩进、空格和换行）...")
-                prompt_lines.append("<<END>>")
-                prompt_lines.append("<<NEW>>")
-                prompt_lines.append("...新内容，可留空表示清空，注意保持结构完整...")
-                prompt_lines.append("<<END>>")
-                prompt_lines.append("[/replace]")
-                prompt_lines.append("<<<END_MODIFY>>>")
-                prompt_lines.append("请勿在普通回复中直接粘贴补丁，必须通过 modify_file 工具发送。")
-                follow_prompt = "\n".join(prompt_lines)
-                messages.append({
-                    "role": "system",
-                    "content": follow_prompt
-                })
-                web_terminal.context_manager.add_conversation("system", follow_prompt)
-                debug_log("已注入修改任务提示")
-
-            if modify_result["handled"] and modify_result.get("failed_blocks"):
-                mark_force_thinking(web_terminal, reason="modify_partial_failure")
-            if modify_result["handled"] and modify_result.get("forced") and modify_result.get("success"):
-                mark_force_thinking(web_terminal, reason="modify_forced_finish")
-            if modify_result["handled"] and not modify_result.get("success"):
-                error_message = modify_result.get("summary_message") or modify_result.get("error") or "修改操作未成功，请根据提示重新执行。"
-                sender('system_message', {
-                    'content': f'⚠️ 修改操作存在未完成的内容：{error_message}'
-                })
-                maybe_mark_failure_from_message(web_terminal, f'⚠️ 修改操作存在未完成的内容：{error_message}')
-                mark_force_thinking(web_terminal, reason="modify_failed")
-            
         if web_terminal.api_client.last_call_used_thinking and current_thinking:
             web_terminal.api_client.current_task_thinking = current_thinking or ""
         if web_terminal.api_client.current_task_first_call:
@@ -885,7 +732,7 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
         update_thinking_after_call(web_terminal)
         
         # 检测是否有格式错误的工具调用
-        if not tool_calls and full_response and AUTO_FIX_TOOL_CALL and not append_result["handled"] and not modify_result["handled"]:
+        if not tool_calls and full_response and AUTO_FIX_TOOL_CALL:
             if detect_malformed_tool_call(full_response):
                 auto_fix_attempts += 1
                 
@@ -919,10 +766,6 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
         
         if full_response:
             assistant_content_parts.append(full_response)
-        elif append_result["handled"] and append_result["assistant_content"]:
-            assistant_content_parts.append(append_result["assistant_content"])
-        elif modify_result["handled"] and modify_result.get("assistant_content"):
-            assistant_content_parts.append(modify_result["assistant_content"])
         
         assistant_content = "\n".join(assistant_content_parts) if assistant_content_parts else ""
         
@@ -949,35 +792,6 @@ async def handle_task_with_sender(terminal: WebTerminal, workspace: UserWorkspac
         text_started = False
         text_has_content = False
         full_response = ""
-        
-        if append_result["handled"] and append_result.get("tool_content"):
-            tool_call_id = append_result.get("tool_call_id") or f"append_{int(time.time() * 1000)}"
-            system_notice = format_tool_result_notice("append_to_file", tool_call_id, append_result["tool_content"])
-            messages.append({
-                "role": "system",
-                "content": system_notice
-            })
-            append_result["tool_call_id"] = tool_call_id
-            debug_log("已将 append_to_file 工具结果以 system 形式追加到对话上下文")
-        if modify_result["handled"] and modify_result.get("tool_content"):
-            tool_call_id = modify_result.get("tool_call_id") or f"modify_{int(time.time() * 1000)}"
-            system_notice = format_tool_result_notice("modify_file", tool_call_id, modify_result["tool_content"])
-            messages.append({
-                "role": "system",
-                "content": system_notice
-            })
-            modify_result["tool_call_id"] = tool_call_id
-            debug_log("已将 modify_file 工具结果以 system 形式追加到对话上下文")
-        
-        force_continue = append_result["handled"] or modify_result["handled"]
-        if force_continue:
-            if append_result["handled"]:
-                debug_log("append_to_file 已处理，继续下一轮以让模型返回确认回复")
-            elif modify_result["handled"]:
-                debug_log("modify_file 已处理，继续下一轮以让模型返回确认回复")
-            else:
-                debug_log("补丁处理完成，继续下一轮以获取模型回复")
-            continue
 
         if not tool_calls:
             debug_log("没有工具调用，结束迭代")
diff --git a/server/state.py b/server/state.py
index aae1e39..0ac7877 100644
--- a/server/state.py
+++ b/server/state.py
@@ -28,7 +28,7 @@ stop_flags: Dict[str, Dict[str, Any]] = {}
 active_polling_tasks: Dict[str, bool] = {}  # conversation_id -> is_polling
 
 # 监控/限流/用量
-MONITOR_FILE_TOOLS = {'append_to_file', 'modify_file', 'write_file', 'edit_file'}
+MONITOR_FILE_TOOLS = {'write_file', 'edit_file'}
 MONITOR_MEMORY_TOOLS = {'update_memory'}
 MONITOR_SNAPSHOT_CHAR_LIMIT = 60000
 MONITOR_MEMORY_ENTRY_LIMIT = 256
diff --git a/static/src/app.ts b/static/src/app.ts
index 9f8120b..0cfb11c 100644
--- a/static/src/app.ts
+++ b/static/src/app.ts
@@ -94,8 +94,6 @@ const appOptions = {
                 chatAppendTextChunk: 'appendTextChunk',
                 chatCompleteTextAction: 'completeText',
                 chatAddSystemMessage: 'addSystemMessage',
-                chatAddAppendPayloadAction: 'addAppendPayloadAction',
-                chatAddModifyPayloadAction: 'addModifyPayloadAction',
                 chatEnsureAssistantMessage: 'ensureAssistantMessage'
             }),
             ...mapActions(useInputStore, {
diff --git a/static/src/app/methods/history.ts b/static/src/app/methods/history.ts
index 54eccdf..9558d8c 100644
--- a/static/src/app/methods/history.ts
+++ b/static/src/app/methods/history.ts
@@ -202,47 +202,8 @@ export const historyMethods = {
                 }
 
                 // 处理普通文本内容（移除思考标签后的内容）
-                const metadata = message.metadata || {};
-                const appendPayloadMeta = metadata.append_payload;
-                const modifyPayloadMeta = metadata.modify_payload;
-                const isAppendMessage = message.name === 'append_to_file';
-                const isModifyMessage = message.name === 'modify_file';
-                const containsAppendMarkers = /<<<\s*(APPEND|MODIFY)/i.test(content || '') || /<<<END_\s*(APPEND|MODIFY)>>>/i.test(content || '');
-
                 const textContent = content.trim();
-
-                if (appendPayloadMeta) {
-                    currentAssistantMessage.actions.push({
-                        id: `history-append-payload-${Date.now()}-${Math.random()}`,
-                        type: 'append_payload',
-                        append: {
-                            path: appendPayloadMeta.path || '未知文件',
-                            forced: !!appendPayloadMeta.forced,
-                            success: appendPayloadMeta.success === undefined ? true : !!appendPayloadMeta.success,
-                            lines: appendPayloadMeta.lines ?? null,
-                            bytes: appendPayloadMeta.bytes ?? null
-                        },
-                        timestamp: Date.now()
-                    });
-                    debugLog('添加append占位信息:', appendPayloadMeta.path);
-                } else if (modifyPayloadMeta) {
-                    currentAssistantMessage.actions.push({
-                        id: `history-modify-payload-${Date.now()}-${Math.random()}`,
-                        type: 'modify_payload',
-                        modify: {
-                            path: modifyPayloadMeta.path || '未知文件',
-                            total: modifyPayloadMeta.total_blocks ?? null,
-                            completed: modifyPayloadMeta.completed || [],
-                            failed: modifyPayloadMeta.failed || [],
-                            forced: !!modifyPayloadMeta.forced,
-                            details: modifyPayloadMeta.details || []
-                        },
-                        timestamp: Date.now()
-                    });
-                    debugLog('添加modify占位信息:', modifyPayloadMeta.path);
-                }
-
-                if (textContent && !appendPayloadMeta && !modifyPayloadMeta && !isAppendMessage && !isModifyMessage && !containsAppendMarkers) {
+                if (textContent) {
                     currentAssistantMessage.actions.push({
                         id: `history-text-${Date.now()}-${Math.random()}`,
                         type: 'text',
@@ -341,12 +302,7 @@ export const historyMethods = {
                                 toolAction.tool.message = result.message;
                             }
                         }
-                        if (message.name === 'append_to_file' && result && result.message) {
-                            toolAction.tool.message = result.message;
-                        }
                         debugLog(`更新工具结果: ${message.name} -> ${message.content?.substring(0, 50)}...`);
-
-                        // append_to_file 的摘要在 append_payload 占位中呈现，此处无需重复
                     } else {
                         console.warn('找不到对应的工具调用:', message.name, message.tool_call_id);
                     }
diff --git a/static/src/components/chat/actions/ToolAction.vue b/static/src/components/chat/actions/ToolAction.vue
index 76227dd..c123aeb 100644
--- a/static/src/components/chat/actions/ToolAction.vue
+++ b/static/src/components/chat/actions/ToolAction.vue
@@ -105,10 +105,6 @@ function renderEnhancedToolResult(): string {
     return renderWriteFile(result, args);
   } else if (name === 'edit_file') {
     return renderEditFile(result, args);
-  } else if (name === 'append_to_file') {
-    return renderAppendToFile(result, args);
-  } else if (name === 'modify_file') {
-    return renderModifyFile(result, args);
   } else if (name === 'delete_file') {
     return renderDeleteFile(result, args);
   } else if (name === 'rename_file') {
@@ -297,67 +293,6 @@ function renderEditFile(result: any, args: any): string {
   return html;
 }
 
-function renderAppendToFile(result: any, args: any): string {
-  const path = args.path || result.path || '';
-  const status = result.success ? '✓ 已追加' : '✗ 失败';
-  const content = args.content || '';
-
-  let html = '<div class="tool-result-meta">';
-  html += `<div><strong>路径：</strong>${escapeHtml(path)}</div>`;
-  html += `<div><strong>状态：</strong>${status}</div>`;
-  html += '</div>';
-
-  if (content) {
-    html += '<div class="tool-result-content scrollable">';
-    html += '<div class="content-label">追加内容：</div>';
-    html += `<pre>${escapeHtml(content)}</pre>`;
-    html += '</div>';
-  }
-
-  return html;
-}
-
-function renderModifyFile(result: any, args: any): string {
-  const path = args.path || result.path || '';
-  const status = result.success ? '✓ 已修改' : '✗ 失败';
-  const operations = args.operations || [];
-
-  let html = '<div class="tool-result-meta">';
-  html += `<div><strong>路径：</strong>${escapeHtml(path)}</div>`;
-  html += `<div><strong>状态：</strong>${status}</div>`;
-  html += '</div>';
-
-  if (operations.length > 0) {
-    html += '<div class="tool-result-diff scrollable">';
-    operations.forEach((op: any, idx: number) => {
-      if (idx > 0) html += '<div class="diff-separator">⋮</div>';
-
-      const type = op.type || op.operation;
-      const oldText = op.old_text || op.old || '';
-      const newText = op.new_text || op.new || '';
-
-      html += `<div class="diff-operation">[${idx + 1}] ${escapeHtml(type)}</div>`;
-
-      if (oldText) {
-        const oldLines = oldText.split('\n');
-        oldLines.forEach((line: string) => {
-          html += `<div class="diff-line diff-remove">- ${escapeHtml(line)}</div>`;
-        });
-      }
-
-      if (newText) {
-        const newLines = newText.split('\n');
-        newLines.forEach((line: string) => {
-          html += `<div class="diff-line diff-add">+ ${escapeHtml(line)}</div>`;
-        });
-      }
-    });
-    html += '</div>';
-  }
-
-  return html;
-}
-
 function renderDeleteFile(result: any, args: any): string {
   const path = args.path || result.path || '';
   const status = result.success ? '✓ 已删除' : '✗ 失败';
diff --git a/static/src/components/chat/actions/toolRenderers.ts b/static/src/components/chat/actions/toolRenderers.ts
index 7e9646a..9edaa18 100644
--- a/static/src/components/chat/actions/toolRenderers.ts
+++ b/static/src/components/chat/actions/toolRenderers.ts
@@ -45,10 +45,6 @@ export function renderEnhancedToolResult(
     return renderWriteFile(result, args);
   } else if (name === 'edit_file') {
     return renderEditFile(result, args);
-  } else if (name === 'append_to_file') {
-    return renderAppendToFile(result, args);
-  } else if (name === 'modify_file') {
-    return renderModifyFile(result, args);
   } else if (name === 'delete_file') {
     return renderDeleteFile(result, args);
   } else if (name === 'rename_file') {
@@ -242,79 +238,6 @@ function renderEditFile(result: any, args: any): string {
   return html;
 }
 
-function renderAppendToFile(result: any, args: any): string {
-  const path = args.file_path || args.path || result.path || '';
-  const status = result.success ? '✓ 已追加' : '✗ 失败';
-  let content = args.content || '';
-
-  // 处理转义的 \n
-  content = content.replace(/\\n/g, '\n');
-
-  let html = '<div class="tool-result-meta">';
-  html += `<div><strong>路径：</strong>${escapeHtml(path)}</div>`;
-  html += `<div><strong>状态：</strong>${status}</div>`;
-  html += '</div>';
-
-  if (content) {
-    html += '<div class="tool-result-content scrollable">';
-    html += '<div class="content-label">追加内容：</div>';
-    const lines = content.split('\n');
-    html += '<div class="tool-result-diff">';
-    lines.forEach((line: string) => {
-      html += `<div class="diff-line diff-add">+ ${escapeHtml(line)}</div>`;
-    });
-    html += '</div>';
-    html += '</div>';
-  }
-
-  return html;
-}
-
-function renderModifyFile(result: any, args: any): string {
-  const path = args.file_path || args.path || result.path || '';
-  const status = result.success ? '✓ 已修改' : '✗ 失败';
-  const operations = args.operations || [];
-
-  let html = '<div class="tool-result-meta">';
-  html += `<div><strong>路径：</strong>${escapeHtml(path)}</div>`;
-  html += `<div><strong>状态：</strong>${status}</div>`;
-  html += '</div>';
-
-  if (operations.length > 0) {
-    html += '<div class="tool-result-diff scrollable">';
-    operations.forEach((op: any, idx: number) => {
-      if (idx > 0) html += '<div class="diff-separator">⋮</div>';
-
-      const type = op.type || op.operation;
-      let oldText = op.old_text || op.old || '';
-      let newText = op.new_text || op.new || '';
-
-      // 处理转义的 \n
-      oldText = oldText.replace(/\\n/g, '\n');
-      newText = newText.replace(/\\n/g, '\n');
-
-      html += `<div class="diff-operation">[${idx + 1}] ${escapeHtml(type)}</div>`;
-
-      if (oldText) {
-        const oldLines = oldText.split('\n');
-        oldLines.forEach((line: string) => {
-          html += `<div class="diff-line diff-remove">- ${escapeHtml(line)}</div>`;
-        });
-      }
-
-      if (newText) {
-        const newLines = newText.split('\n');
-        newLines.forEach((line: string) => {
-          html += `<div class="diff-line diff-add">+ ${escapeHtml(line)}</div>`;
-        });
-      }
-    });
-    html += '</div>';
-  }
-
-  return html;
-}
-
 function renderDeleteFile(result: any, args: any): string {
   const path = args.path || result.path || '';
   const status = result.success ? '✓ 已删除' : '✗ 失败';
diff --git a/static/src/components/personalization/PersonalizationDrawer.vue b/static/src/components/personalization/PersonalizationDrawer.vue
index ef57583..d426859 100644
--- a/static/src/components/personalization/PersonalizationDrawer.vue
+++ b/static/src/components/personalization/PersonalizationDrawer.vue
@@ -517,7 +517,30 @@
                           ></path>
                         </svg>
                       </span>
-                      <span>在工具块显示“我要做什么”的简短提示</span>
+                      <span>在工具块显示"我要做什么"的简短提示</span>
+                    </label>
+                  </div>
+                  <div class="behavior-field">
+                    <div class="behavior-field-header">
+                      <span class="field-title">Skill 提示系统</span>
+                      <p class="field-desc">开启后，检测到特定关键词时会自动提示模型阅读相关 skill（如 terminal-guide、sub-agent-guide 等）。</p>
+                    </div>
+                    <label class="toggle-row">
+                      <input
+                        type="checkbox"
+                        :checked="form.skill_hints_enabled"
+                        @change="personalization.updateField({ key: 'skill_hints_enabled', value: $event.target.checked })"
+                      />
+                      <span class="fancy-check" aria-hidden="true">
+                        <svg viewBox="0 0 64 64">
+                          <path
+                            d="M 0 16 V 56 A 8 8 90 0 0 8 64 H 56 A 8 8 90 0 0 64 56 V 8 A 8 8 90 0 0 56 0 H 8 A 8 8 90 0 0 0 8 V 16 L 32 48 L 64 16 V 8 A 8 8 90 0 0 56 0 H 8 A 8 8 90 0 0 0 8 V 56 A 8 8 90 0 0 8 64 H 56 A 8 8 90 0 0 64 56 V 16"
+                            pathLength="575.0541381835938"
+                            class="fancy-path"
+                          ></path>
+                        </svg>
+                      </span>
+                      <span>根据关键词自动提示阅读相关 skill</span>
                     </label>
                   </div>
                   <div class="behavior-field">
diff --git a/static/src/stores/chat.ts b/static/src/stores/chat.ts
index 37e9f95..21adc6a 100644
--- a/static/src/stores/chat.ts
+++ b/static/src/stores/chat.ts
@@ -286,26 +286,6 @@ export const useChatStore = defineStore('chat', {
         timestamp: Date.now()
       });
     },
-    addAppendPayloadAction(data: any) {
-      const msg = this.ensureAssistantMessage();
-      clearAwaitingFirstContent(msg);
-      msg.actions.push({
-        id: `append-payload-${Date.now()}-${Math.random()}`,
-        type: 'append_payload',
-        append: data,
-        timestamp: Date.now()
-      });
-    },
-    addModifyPayloadAction(data: any) {
-      const msg = this.ensureAssistantMessage();
-      clearAwaitingFirstContent(msg);
-      msg.actions.push({
-        id: `modify-payload-${Date.now()}-${Math.random()}`,
-        type: 'modify_payload',
-        modify: data,
-        timestamp: Date.now()
-      });
-    },
     getActiveThinkingAction(msg: any) {
       if (!msg || !Array.isArray(msg.actions)) {
         return null;
diff --git a/static/src/stores/monitor.ts b/static/src/stores/monitor.ts
index 47c6363..1dd863b 100644
--- a/static/src/stores/monitor.ts
+++ b/static/src/stores/monitor.ts
@@ -101,8 +101,6 @@ const TOOL_SCENE_MAP: Record<string, string> = {
   create_file: 'createFile',
   rename_file: 'renameFile',
   delete_file: 'deleteFile',
-  append_to_file: 'appendFile',
-  modify_file: 'modifyFile',
   write_file: 'modifyFile',
   edit_file: 'modifyFile',
   run_command: 'runCommand',
diff --git a/static/src/stores/personalization.ts b/static/src/stores/personalization.ts
index b1ed19b..59c87b3 100644
--- a/static/src/stores/personalization.ts
+++ b/static/src/stores/personalization.ts
@@ -6,6 +6,7 @@ interface PersonalForm {
   enabled: boolean;
   auto_generate_title: boolean;
   tool_intent_enabled: boolean;
+  skill_hints_enabled: boolean;
   silent_tool_disable: boolean;
   enhanced_tool_display: boolean;
   enabled_skills: string[];
@@ -63,6 +64,7 @@ const defaultForm = (): PersonalForm => ({
   enabled: false,
   auto_generate_title: true,
   tool_intent_enabled: true,
+  skill_hints_enabled: false,
   silent_tool_disable: false,
   enhanced_tool_display: true,
   enabled_skills: [],
@@ -196,6 +198,7 @@ export const usePersonalizationStore = defineStore('personalization', {
         enabled: !!data.enabled,
         auto_generate_title: data.auto_generate_title !== false,
         tool_intent_enabled: !!data.tool_intent_enabled,
+        skill_hints_enabled: !!data.skill_hints_enabled,
         silent_tool_disable: !!data.silent_tool_disable,
         enhanced_tool_display: data.enhanced_tool_display !== false,
         enabled_skills: Array.isArray(data.enabled_skills)
diff --git a/static/src/utils/chatDisplay.ts b/static/src/utils/chatDisplay.ts
index 0073142..7ee6ab0 100644
--- a/static/src/utils/chatDisplay.ts
+++ b/static/src/utils/chatDisplay.ts
@@ -9,8 +9,6 @@ const RUNNING_ANIMATIONS: Record<string, string> = {
   rename_file: 'file-animation',
   write_file: 'file-animation',
   edit_file: 'file-animation',
-  modify_file: 'file-animation',
-  append_to_file: 'file-animation',
   create_folder: 'file-animation',
   web_search: 'search-animation',
   extract_webpage: 'search-animation',
@@ -36,8 +34,6 @@ const RUNNING_STATUS_TEXTS: Record<string, string> = {
   rename_file: '正在重命名文件...',
   write_file: '正在写入文件...',
   edit_file: '正在编辑文件...',
-  modify_file: '正在修改文件...',
-  append_to_file: '正在追加文件...',
   create_folder: '正在创建文件夹...',
   web_search: '正在搜索网络...',
   extract_webpage: '正在提取网页...',
@@ -57,8 +53,6 @@ const COMPLETED_STATUS_TEXTS: Record<string, string> = {
   rename_file: '文件重命名成功',
   write_file: '文件写入完成',
   edit_file: '文件编辑完成',
-  modify_file: '文件修改成功',
-  append_to_file: '文件追加完成',
   create_folder: '文件夹创建成功',
   web_search: '搜索完成',
   extract_webpage: '网页提取完成',
diff --git a/static/src/utils/icons.ts b/static/src/utils/icons.ts
index 21dfcfb..c877de9 100644
--- a/static/src/utils/icons.ts
+++ b/static/src/utils/icons.ts
@@ -41,14 +41,12 @@ export const ICONS = Object.freeze({
 });
 
 export const TOOL_ICON_MAP = Object.freeze({
-  append_to_file: 'pencil',
   close_sub_agent: 'bot',
   create_file: 'file',
   create_folder: 'folder',
   create_sub_agent: 'bot',
   delete_file: 'trash',
   extract_webpage: 'globe',
-  modify_file: 'pencil',
   write_file: 'pencil',
   edit_file: 'pencil',
   vlm_analyze: 'camera',
diff --git a/utils/api_client.py b/utils/api_client.py
index 3700e41..a38c936 100644
--- a/utils/api_client.py
+++ b/utils/api_client.py
@@ -831,9 +831,6 @@ class DeepSeekClient:
             full_response = ""
             tool_calls = []
             current_thinking = ""
-            # 针对 append_to_file / modify_file 的占位结构，防止未定义变量导致异常
-            append_result = {"handled": False}
-            modify_result = {"handled": False}
             
             # 状态标志
             in_thinking = False
@@ -935,10 +932,6 @@ class DeepSeekClient:
             # 添加正式回复内容（如果有）
             if full_response:
                 assistant_content_parts.append(full_response)
-            elif append_result["handled"] and append_result["assistant_content"]:
-                assistant_content_parts.append(append_result["assistant_content"])
-            elif modify_result["handled"] and modify_result.get("assistant_content"):
-                assistant_content_parts.append(modify_result["assistant_content"])
             
             # 添加工具调用说明
             if tool_calls:
@@ -1000,31 +993,6 @@ class DeepSeekClient:
                 
                 self._print(f"\n{OUTPUT_FORMATS['action']} 调用工具: {function_name}")
                 
-                # 额外的参数长度检查（针对特定工具）
-                if function_name == "modify_file" and "content" in arguments:
-                    content_length = len(arguments.get("content", ""))
-                    if content_length > 9999999999:  # 降低到50KB限制
-                        error_msg = f"内容过长({content_length}字符)，超过50KB限制"
-                        self._print(f"{OUTPUT_FORMATS['warning']} {error_msg}")
-                        
-                        messages.append({
-                            "role": "tool",
-                            "tool_call_id": tool_call["id"],
-                            "name": function_name,
-                            "content": json.dumps({
-                                "success": False,
-                                "error": error_msg,
-                                "suggestion": "请将内容分成多个小块分别修改，或使用replace操作只修改必要部分"
-                            }, ensure_ascii=False)
-                        })
-                        
-                        all_tool_results.append({
-                            "tool": function_name,
-                            "args": arguments,
-                            "result": error_msg
-                        })
-                        continue
-                
                 tool_result = await tool_handler(function_name, arguments)
                 
                 # 解析工具结果，提取关键信息