dsh-data-cleaning-agent 0.9.15 → 0.9.17

package/CHANGELOG.md

@@ -1,3 +1,16 @@
+# 0.9.17 · 2026-09-18
+
+- 工商与简介新增 14 项一企一行字段：省市区及地区代码、四级国标行业、四级企查查行业、主营产品、企业规模；不改变默认选择。
+- 兼容行业由文本改为对象的返回结构，保留旧文本且不推测缺失层级，避免输出对象字符串；主营产品按源顺序合并一格。
+- 向导与工作台同步字段目录、别名搜索、层级/规模说明及动态数量，保留原列回填和客户结果/任务报告分层。
+
+# 0.9.16 · 2026-09-15
+
+- UX-49：仅清洗入口新建的独立 Session，通过 Host 原生草稿接口预填一次；空草稿、附件/引用、输入就绪、IME、草稿版本与晚到竞态二次核验，不自动发送、建业务任务或展开工作台。
+- 用户编辑/主动清空、旧会话、刷新、重挂载和切回不补回、不清理草稿；保留 UX-48 接纳后开台与 settling→hero 品牌修复。
+- 初始引导采用稳定模板 ID/版本和精确文本指纹；向导排除未改模板，修改后的内容不整体忽略，已有草稿提供替换/追加/取消，异步回填前重新核验。
+- 未补占位符或缺名单/待补字段，在清洗业务 Skill 中优先澄清，不调 OAuth/OCR/MCP 或创建 workflow；不拦截 DSH 全局发送。
+
 # 0.9.15 · 2026-09-14
 
 - 保留 0.9.14 的 Host 驱动单向工作流、导航保护、导入核验和紧凑状态区功能。

package/README.en.md

@@ -10,7 +10,11 @@ Related agents: [数据清洗补全](https://github.com/duhu2000/dsh-data-cleani
 
 > A data cleaning & completion agent plugin for DeepSeek Harness: local CSV/XLSX/JSON engine plus optional Qichacha (QCC) MCP enterprise-data enrichment. Initiated and maintained by the Qichacha (QCC) team.
 >
-> Current source version / 当前源码版本: **0.9.15** (stable release)
+> Current source version / 当前源码版本: **0.9.17** (stable release)
+
+0.9.17 adds 14 region, industry-level, product and enterprise-size fields (150 total), supports both legacy industry text and nested responses, and improves field search and column mapping. Defaults remain unchanged; missing levels are never inferred. Products reflect only the upstream first 10 entries. See [field expansion and verification boundaries](docs/CORE-FIELDS-20260918.md).
+
+0.9.16 implements UX-49: a one-time editable native guide for new cleaning-entry Sessions, with existing text, attachment, IME and asynchronous-race guards. Clearing, refreshing or returning never restores it. Initialization does not send, call tools or open the workbench. Incomplete templates request the missing list and fields first. See the [adoption record](docs/UNIFIED-HOME-1.5.4-ADOPTION.md) for isolated DSH evidence and unverified boundaries.
 
 0.9.15 releases the Host-driven forward-only workflow across import, rules, entity matching, enrichment, and download. Future stages provide explicit locked feedback, while completed stages remain viewable without business-state mutations. Imported rows stay visible for full review before the bottom confirmation advances to rules. The compact status area reports stage, source, time, scope, and explicit result counts. The 0.9.13 submission-admission, background-session isolation, and manual-collapse guarantees remain intact. Start a new cleaning Session after upgrading; legacy Session migration is outside this release's acceptance scope.
 

package/README.md

@@ -9,7 +9,7 @@
 数据清洗补全智能体：面向 Excel/CSV/JSON 企业名单，提供数据清洗、表格清洗、清洗补全、去重、企业数据补全与字段补全，支持企查查 MCP 和结果导出。
 
 ```sh
-dsh plugin --profile web add dsh-data-cleaning-agent@0.9.15
+dsh plugin --profile web add dsh-data-cleaning-agent@0.9.17
 ```
 
 请先满足下文的 DSH、连接器及侧边栏依赖要求；安装后完整停止并重启对应 Profile。
@@ -27,7 +27,11 @@ dsh plugin --profile web add dsh-data-cleaning-agent@0.9.15
 
 > 在 DeepSeek Harness 中清洗、补全、画像企业名单数据的智能体插件：本地 CSV/XLSX/JSON 引擎 + 可选企查查（Qichacha/QCC）MCP 企业数据补全，由企查查（Qichacha/QCC）团队发起并维护。
 >
-> 当前源码版本 / Current source version: **0.9.15**（正式版本）
+> 当前源码版本 / Current source version: **0.9.17**（正式版本）
+
+0.9.17 新增省市区、地区代码、四级国标行业、四级企查查行业、主营产品与企业规模共 14 项字段，目录共 150 项；兼容行业由文本改为分层对象，优化字段搜索与原列映射。默认选择不扩大，缺失层级不推测；主营产品最多为上游返回的前 10 项。详见[字段扩展与验证边界](docs/CORE-FIELDS-20260918.md)。
+
+0.9.16 实施 UX-49：首页仅为清洗入口新建的会话预填一次原生可编辑引导，保护已有文字、附件、中文组合输入与异步竞态。主动清空、刷新或切回不会补回；不自动发送、调用工具或展开工作台。未补全模板先澄清名单和字段。真实 DSH 隔离验收及未验证边界见[规范采用记录](docs/UNIFIED-HOME-1.5.4-ADOPTION.md)。
 
 0.9.15 发布 Host 驱动的单向工作流：统一“导入与核验 → 规则与体检 → 主体匹配 → 字段补全 → 结果下载”五步导航。未来节点给出明确锁定反馈，历史节点保留查看但不挂载业务写操作；导入后先核验完整名单，再从表格底部确认进入规则页。紧凑状态区展示节点、来源、时间、任务范围和明确结果计数。0.9.13 的真实提交接纳、后台会话隔离与手动收起保护保持不变。升级后请新建清洗会话验证，旧 Session 迁移不在本版验收范围。
 

package/docs/COMPATIBILITY.md

@@ -206,3 +206,10 @@ Host 制品下载按钮。alpha.2 继续只检查 Host/路由/制品 Bridge，
   XLSX 因此以 Base64 文本落盘。未来切换二进制 seam 必须保留旧 `v1` 制品读取兼容。
 - v2 已生成制品可跨 Host 重启恢复，但浏览器 runtime 中尚未导出的原始行不会持久化；恢复中途任务
   仍需用户重新提供输入。制品只在当前工作区，不承诺跨设备同步。
+# UX-49 原生初始草稿兼容边界（0.9.16，2026-09-15）
+
+- 采用记录见 `UNIFIED-HOME-1.5.4-ADOPTION.md` 的 v1.5.6 小节。仅新建清洗 Session 初始化，旧 Session/刷新/重挂载只恢复归属，不迁移或补回草稿。
+- 本机核对 DSH 0.1.2-rc.1 的公开 `conversation.input.shell(sessionId)`：`snapshot.draft / draftRev / phase / imageIds / occurrences` 与 `setDraft`。缺失任一必要快照能力时保守跳过，不使用 DOM 空文本推断可写，不读取私有 editor。
+- Host InputState 没有 IME 字段：使用 compositionstart/end 与入口异步期间的 beforeinput/input/paste/drop 事件，仅用于取消本插件初始化，不阻断默认事件、不修改其他产品。公开 shell 可在新 Session 的输入 DOM 挂载前使用；不把 hero/settling 或 session.blank 当草稿就绪证据。
+- 未填写的模板发送由本业务 Skill 澄清；不改 Host submit/beginSubmission 全局语义。UX-48 原有 observed 接纳桥保持独立，草稿写入不产生业务接纳事件。
+- 本轮不升级 DSH、不修改正式 Profile、不请求 QCC/OCR/模型凭据。无法验证的 Host 版本或真实模型决策明确保留为外部边界。

package/docs/CORE-FIELDS-20260918.md

@@ -0,0 +1,50 @@
+# V3.25 核心字段扩展评估与实现（0.9.17，2026-09-18）
+
+## 一手依据
+
+- 2026-09-17 V3.25.0 发版通知及核对说明：工商地区/国标行业、简介企查查行业/主营产品/企业规模。
+- qcc_mcp_new 本地 dev@fabceff：apps/mcp/src/tools/company_registration_info.tool.ts、company_profile.tool.ts；工商变更 cf4cfe1，简介 9203e20/f9719a0/a538e16/3b672ec。
+- mcp_web 本地 dev@f2b85a79：apps/web/lib/demo-data.ts 及两个对应 example tests；913882d4、c9fb123a 和 109c9216。两个源仓只读，未修改工作树。
+- 清洗插件基线 origin/main@794201b / npm latest 0.9.16，无在途 PR。独立分支 feat/core-fields-20260918；不变更共享字段包或默认选择。
+- 发版核对说明明确未独立验证生产部署；本次亦未发起收费 MCP 请求。代码/官网示例对齐不能替代线上验收。
+
+## 新增 14 项一对一字段
+
+| 来源工具 | 精确返回路径 | 补全字段 ID | 显示名 |
+| --- | --- | --- | --- |
+| get_company_registration_info | 地区信息.省份 | province | 省份 |
+| 同上 | 地区信息.城市 | city | 城市 |
+| 同上 | 地区信息.区域 | district | 区域 |
+| 同上 | 地区信息.地区代码 | area_code | 地区代码 |
+| 同上 | 国标行业.门类 | industry_section | 国标行业门类 |
+| 同上 | 国标行业.大类 | industry_large | 国标行业大类 |
+| 同上 | 国标行业.中类 | industry_middle | 国标行业中类 |
+| 同上 | 国标行业.小类 | industry_small | 国标行业小类 |
+| get_company_profile | 企查查行业.一级 | qcc_industry_level1 | 企查查行业一级 |
+| 同上 | 企查查行业.二级 | qcc_industry_level2 | 企查查行业二级 |
+| 同上 | 企查查行业.三级 | qcc_industry_level3 | 企查查行业三级 |
+| 同上 | 企查查行业.四级 | qcc_industry_level4 | 企查查行业四级 |
+| 同上 | 主营产品 | main_products | 主营产品 |
+| 同上 | 企业规模 | company_scale | 企业规模 |
+
+两个工具均来自 qcc-company。目录 136→150；工商 27→35、简介 3→9。仍按企业+工具去重调用，不为每个层级重复查同一工具。
+
+## 兼容性与数据边界
+
+- 旧 industry_category / qcc_industry ID 保留：旧文本原样输出；新对象按明确层级标注串联，禁止直接 String(object)。独立层级只取对应叶子，不把旧文本认定为某层级，也不从其他层级补缺。
+- 地区字段来自嵌套“地区信息”，不从所属地区/注册地址拆分；地区代码保留字符串与前导零。不同主体可能不返回这些字段，缺失留空。
+- 企业规模取简介“企业规模”（上游大型/中型/小型/微型）；原 company_size 仍是工商“人员规模”，不能互相回填。
+- 主营产品是名称数组，按原顺序以中文分号合并单元格，不展开企业行、不重排、不去重。当前服务只请求前 10 项，空数组也可能来自不支持主体或补充接口异常，不能据此判断企业没有产品。
+- 原表非空值保护、只回填选定字段、来源审计和客户 XLSX/任务报告分层保持不变；主结果不增加本次未选字段。
+
+## 选择界面
+
+保持两大来源分组及现有全选/清空交互；增加层级、规模和产品边界说明。新增省/市/区县、区域编码、企查查一级行业等别名，搜索和表头映射共用；不把笼统“行业”自动绑定到某级。数量取当前目录，移除过时固定批次数字。默认字段不扩大。
+
+## 验证
+
+自动化覆盖：新结构/旧文本、缺层级、畸形对象、零值和前导零、产品原顺序、两个工具去重、原值保留、未选择字段不输出、前后端目录映射及 UI 数量/搜索。
+
+本地验收：npm run check 全部通过（302 项测试及语法、文档、打包门禁）；真实 React/Chromium 亮/暗主题 × 5 种尺寸共 10 组回归通过，包括新增别名搜索、行业层级、银行模板省市区自动映射。XLSX 合成往返测试验证中文列名、前导零、产品单元格和报告分层；未在客户已登录的 3080 会话上发起业务调用。
+
+未验证：生产 MCP 实际部署与权限、真实企业返回完整度；不将合成 fixture 当作真实业务验收。用户后续已授权 commit、push、tag 与 npm 发布，0.9.17 按 PR 检查、合并默认分支、tag/OIDC 流程发布；发布状态以 registry 回读为准。

package/docs/UI-WORKFLOW-V2.md

@@ -18,14 +18,15 @@
 
 ## 2. 当前范围
 
-当前字段目录共 128 项，UI/Host/XLSX 按 8 个来源工具分组：
+当前字段目录共 150 项，UI/Host/XLSX 按 11 个来源工具分组（0.9.17，2026-09-18 扩展）：
 
-- 基础 30 项：`get_company_registration_info`（企业工商信息，27 项）和 `get_company_profile`（企业简介，3 项）。
+- 基础 44 项：`get_company_registration_info`（企业工商信息，35 项）和 `get_company_profile`（企业简介，9 项）。
+- 实际控制人 4 项、受益所有人首条 1 项、主要财务指标 3 项，沿用共享字段契约。
 - 第一批 40 项：`get_contact_info`（联系方式，6 项）、`get_listing_info`（上市信息，15 项）、`get_tax_invoice_info`（税务开票，8 项）、`get_import_export_credit`（进出口信用，11 项）。
 - 第二批 58 项：`get_company_risk_scan`（企业自身风险扫描，38 项）和 `get_company_related_risk_scan`（企业关联风险扫描，20 项）。
 - 基础工商：企业名称、统一社会信用代码、注册号、组织机构代码、纳税人识别号、登记状态、法定代表人、注册/实缴资本、成立日期、企业类型、核准日期、登记机关、纳税人资质、支付系统行号、进出口企业代码、企业简称、英文名。
-- 地址与地区：注册地址、通信地址、所属地区原值。
-- 经营信息：经营范围、国标行业、企查查行业、营业期限、人员规模、参保人数、分支机构参保人数、企业简介、产业链概览。
+- 地址与地区：注册地址、通信地址、所属地区原值及明确返回的省份、城市、区域、地区代码；不拆地址推测。
+- 经营信息：经营范围、国标行业/企查查行业展示值及各自四级独立字段、主营产品、企业规模、营业期限、人员规模、参保人数、分支机构参保人数、企业简介、产业链概览。企业规模与人员规模独立；主营产品按源顺序合并一格，当前上游最多 10 项，并非完整产品库。详见 [字段扩展核对](CORE-FIELDS-20260918.md)。
 - 上传列“联系电话”仍可作为本地质量检查字段；QCC 联系方式另以“首选联系电话/邮箱/官网”等字段输出。电话、邮箱和网址全集不拼接进单元格。
 - 风险字段只输出扫描工具已返回的聚合计数、命中摘要与重点维度命中关联方数；风险明细、关联方名单和下钻结果不进入一企一行主表。
 - 一级行业、二级行业、省、市、区县、曾用名和知识产权摘要不在当前目录；不得推断、拼接或用其他字段冒充。

package/docs/UNIFIED-HOME-1.5.4-ADOPTION.md

@@ -1,4 +1,25 @@
-# DSH-UX-001 v1.5.4 采用记录
+# DSH-UX-001 首页规范采用记录
+
+## v1.5.6 / UX49-CLEAN（2026-09-15，0.9.16）
+
+发布授权补充：用户于实施与 PR 验证后明确要求 commit、push、tag、npm，按仓库既有 CI、合并及 OIDC 发布流程推进 0.9.16；原实施阶段仅授权 PR 的边界作为历史记录保留。
+
+依据共享 `DSH智能体开发交互规范方案.md` 的 UX-49、§11.1、§14，并核对 UX-48；协同清单 `DSH-UX-049-四智能体首页初始引导协同.md` 及总账同日交接。这里只记录采用与差异，不复制规范全文。
+
+- 审计基线：remote main `fc7d57240b843e1a57998cf0ebbe4f39c086befb`，npm latest 0.9.15 / gitHead `6d8b37a7a0efae1aa2fddc72d61fc5e732f3231f`。原仓 main 的未提交 client/test 修改保持原样，另建独立分支。
+- 已有正确项：独立清洗 Session namespace、宿主原生输入接口、实际接纳后开台、标题 settling→hero 修复、历史来源只读与主结果/报告分层。
+- 真实缺陷：入口无条件 setDraft；普通会话异步恢复时按宽泛签名清理草稿；已有业务入口再次点击会开台；向导缺少已有内容确认与晚到写入检查。以上已收紧。
+- 初始模板 `dsh-initial-draft/data-cleaning/1` / version 1，UTF-8 SHA-256 `2b84b86f669e64e970381e4c7ab8547c77d5e10ee7d5441543c893b9f1272e2b`。仅精确相等才排除；不 trim、不按前后缀忽略用户修改。
+- 初始化仅由本入口成功创建的全新 ID 触发，在打开会话前操作未挂载的原生 shell，避免 Host setDraft 的选区写入抢占已挂载输入焦点。snapshot 缺失或非 plain、draft 非空、draftRev 非零、imageIds/occurrences 非空均跳过。建立一次性记录、订阅后重读、微任务后再次快照；期间用户输入/附件/组合事件、会话切换或卸载使写入失效，无轮询补回。
+- 清洗业务 Skill 将信息充分性检查置于任何工具之前：占位符/缺名单或待补字段先澄清；没有注册全局发送拦截。安全凭证所指暂存来源仍可提供名单和范围，不重复要求上传。
+- 向导按 UX-10 提供替换/追加/取消，默认保留用户内容；同一 Session 未改的向导草稿更新而非重复叠加。异步回填重读草稿版本、附件和当前业务归属。
+- 本地门禁：`npm run check` 通过（296 项测试，语法/文档/营销/69 文件 pack 白名单）；Chromium 浅深色 × 5 尺寸 10/10；XLSX 预览 390/800/1440px 3/3。包含附件、IME 事件、晚到修改/清空、卸载、A/B、普通 Session、Tab X/折叠恢复、草稿回填替换/追加/取消与零开台断言。
+- 真实 DSH：`scripts/ux49-native-smoke.mjs`，独立 Profile、无模型/MCP 凭据。DSH 0.1.2-rc.1 + Sidebar 0.18.1 + 清洗候选/招投标 0.5.11/访前尽调 0.1.35/填表 0.2.30 共装通过。原生模板、正确 Hero 标题、不移动焦点、主动清空不补回、刷新保留修改、A/B/普通会话隔离、零业务 API mutation、零可见工作台、零浏览器异常均通过。
+- 实机客户端 SHA-256 `bacfd1faa7eff584ddec3e948ee3f2e19f1d710224de9068d9a62415345a28a1`，测试包 SHA-256 `e93fac1541934a43bc54789283d57c378fbd721fe0d8fc86b922b2a69c5747c5`。原始报告/截图在 `/var/folders/ws/tmsn44b140lf8b89ml5qqjf00000gn/T/cleaning-ux49-native-HtQGy0/`；本地日志 `/private/tmp/ux49-{check,ui,artifact,native}-release-candidate.log`，UI 截图 `_scratch/ui-layout/`。测试包不是 npm 发布物；随后文档和验收脚本路径清理不改变已验收客户端字节。
+- 未验证边界：真实模型执行占位符澄清（仅验证 Skill 优先规则，不冒充实际模型决策）、系统级真实中文输入法（已覆盖组合事件）、收费 QCC/OCR、多账号/其他 Host 版本，以及其余三产品各自 UX49 候选版本。正式 3080/Profile 未修改。
+- 本次仅 commit/push/PR/CI；不合并、不升版本、不 tag/Release/npm publish。
+
+## v1.5.4 历史采用记录
 
 日期：2026-09-14。基线：0.9.11 / a0ec971db0d25d42f67f99e4328cca6f49a2bc2f。
 本轮仅本地代码与验收，不发布，不修改正式 Profile 或全局 DSH。

package/lib/client.js

@@ -1605,19 +1605,104 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       return () => { style.remove(); };
     }
 
-    // 原始数据不进入 Host storageDomain 或模型上下文；只在当前页面按 taskId 隔离。
+    // 原始数据不进入 Host storageDomain 或模型上下文；仅在当前浏览器页签内按 taskId 隔离。
+    // sessionStorage 只用于刷新恢复，关闭页签后不承诺保留；写入失败时降级为同源重新载入。
+    const RUNTIME_SESSION_CACHE_KEY = 'dsh.data-cleaning-agent.runtime-tasks.v1';
+    const RUNTIME_SESSION_CACHE_LIMIT = 20;
+    const RUNTIME_SESSION_CACHE_MAX_ROWS = 5_000;
+    const RUNTIME_SESSION_CACHE_MAX_CHARS = 2 * 1024 * 1024;
     const runtimeTasks = new Map();
     const workflowTaskBySession = new Map();
     const workflowTaskCreationBySession = new Map();
     const workflowOperationBySession = new Map();
 
-    function runtimeFor(taskId, create = true) {
+    function readRuntimeSessionCache() {
+      try {
+        const value = JSON.parse(window.sessionStorage?.getItem(RUNTIME_SESSION_CACHE_KEY) || '{}');
+        return value && typeof value === 'object' && !Array.isArray(value) ? value : {};
+      } catch { return {}; }
+    }
+
+    function runtimeFromSession(taskId, expectedChecksum = '') {
       const key = String(taskId || 'unassigned');
+      const value = readRuntimeSessionCache()[key];
+      const rows = Array.isArray(value?.rows)
+        ? value.rows.filter((row) => row && typeof row === 'object' && !Array.isArray(row))
+        : [];
+      const checksum = String(value?.source?.checksum || '');
+      if (!rows.length || !expectedChecksum || checksum !== expectedChecksum) return null;
+      return {
+        rows,
+        sourceRows: rows,
+        headers: Array.isArray(value.headers) ? value.headers.map(String) : [],
+        source: value.source && typeof value.source === 'object' ? value.source : null,
+        restoredFromSession: true,
+        lastCsv: { clean: null, complete: null, qcc: null, review: null },
+        resultRows: { clean: null, complete: null, qcc: null },
+      };
+    }
+
+    function persistRuntimeSession(taskId, runtime, removeTaskId = null) {
+      const rows = Array.isArray(runtime?.sourceRows) && runtime.sourceRows.length
+        ? runtime.sourceRows
+        : Array.isArray(runtime?.rows) ? runtime.rows : [];
+      if (!rows.length || rows.length > RUNTIME_SESSION_CACHE_MAX_ROWS
+        || typeof window.sessionStorage?.setItem !== 'function') return false;
+      try {
+        const key = String(taskId || 'unassigned');
+        const cache = readRuntimeSessionCache();
+        if (removeTaskId !== null) delete cache[String(removeTaskId || 'unassigned')];
+        delete cache[key];
+        cache[key] = {
+          rows,
+          headers: Array.isArray(runtime.headers) ? runtime.headers : [],
+          source: runtime.source && typeof runtime.source === 'object' ? runtime.source : null,
+        };
+        const entries = Object.entries(cache).slice(-RUNTIME_SESSION_CACHE_LIMIT);
+        while (entries.length) {
+          const serialized = JSON.stringify(Object.fromEntries(entries));
+          if (serialized.length <= RUNTIME_SESSION_CACHE_MAX_CHARS) {
+            window.sessionStorage?.setItem(RUNTIME_SESSION_CACHE_KEY, serialized);
+            return true;
+          }
+          entries.shift();
+        }
+        return false;
+      } catch { return false; }
+    }
+
+    function removeRuntimeSession(taskId) {
+      try {
+        const key = String(taskId || 'unassigned');
+        const cache = readRuntimeSessionCache();
+        if (!(key in cache)) return;
+        delete cache[key];
+        window.sessionStorage?.setItem(RUNTIME_SESSION_CACHE_KEY, JSON.stringify(cache));
+      } catch {}
+    }
+
+    function removeRuntime(taskId) {
+      runtimeTasks.delete(String(taskId || 'unassigned'));
+      removeRuntimeSession(taskId);
+    }
+
+    function runtimeFor(taskId, create = true, expectedChecksum = '') {
+      const key = String(taskId || 'unassigned');
+      const current = runtimeTasks.get(key);
+      if (current && expectedChecksum && current.source?.checksum !== expectedChecksum) {
+        runtimeTasks.delete(key);
+        removeRuntimeSession(key);
+      }
+      if (!runtimeTasks.has(key)) {
+        const restored = runtimeFromSession(key, expectedChecksum);
+        if (restored) runtimeTasks.set(key, restored);
+      }
       if (!runtimeTasks.has(key) && create) {
         runtimeTasks.set(key, {
           rows: [],
           headers: [],
           source: null,
+          restoredFromSession: false,
           lastCsv: { clean: null, complete: null, qcc: null, review: null },
           resultRows: { clean: null, complete: null, qcc: null },
         });
@@ -1633,6 +1718,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       target.sourceRows = from.sourceRows;
       target.headers = from.headers;
       target.source = from.source;
+      target.restoredFromSession = from.restoredFromSession;
       target.lastCsv = from.lastCsv;
       target.resultRows = from.resultRows;
       runtimeTasks.delete(String(fromId || 'unassigned'));
@@ -1714,12 +1800,19 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
         ['taxpayer_qualification', '纳税人资质'], ['payment_line_no', '支付系统行号'],
         ['import_export_company_code', '进出口企业代码'], ['short_name', '企业简称'], ['english_name', '英文名'],
         ['registered_address', '注册地址'], ['mailing_address', '通信地址'], ['region', '所属地区'],
+        ['province', '省份'], ['city', '城市'], ['district', '区域'], ['area_code', '地区代码'],
         ['business_scope', '经营范围'], ['industry_category', '国标行业'],
+        ['industry_section', '国标行业门类'], ['industry_large', '国标行业大类'],
+        ['industry_middle', '国标行业中类'], ['industry_small', '国标行业小类'],
         ['operating_period', '营业期限'], ['company_size', '人员规模'], ['insured_count', '参保人数'],
         ['branch_insured_count', '分支机构参保人数'],
       ], 'get_company_registration_info'],
       ['company_profile', '企业简介', [
-        ['qcc_industry', '企查查行业'], ['company_profile', '企业简介'],
+        ['qcc_industry', '企查查行业'],
+        ['qcc_industry_level1', '企查查行业一级'], ['qcc_industry_level2', '企查查行业二级'],
+        ['qcc_industry_level3', '企查查行业三级'], ['qcc_industry_level4', '企查查行业四级'],
+        ['main_products', '主营产品'], ['company_scale', '企业规模'],
+        ['company_profile', '企业简介'],
         ['industry_chain_overview', '产业链概览'],
       ], 'get_company_profile'],
       ['contact_info', '联系方式', [
@@ -1762,6 +1855,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
     ];
     FIELD_GROUPS.sort((a, b) => FIELD_GROUP_ORDER.indexOf(a[0]) - FIELD_GROUP_ORDER.indexOf(b[0]));
     const ENRICHMENT_OPTIONS = FIELD_GROUPS.flatMap(([, , fields]) => fields);
+    const CORE_FIELD_GUIDE = '可分别选择省市区、地区代码、四级国标行业与四级企查查行业；未分级行业列保留完整层级展示。企业规模不等于人员规模；主营产品按原顺序合并一格（上游最多返回 10 项，非完整产品库）。缺失值留空，不推测。';
     const INPUT_ONLY_MAPPING_OPTIONS = [['phone', '联系电话']];
     const RESULT_FIELD_LABELS = new Map([
       ...ENRICHMENT_OPTIONS,
@@ -1781,7 +1875,12 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
     ];
     const DEFAULT_CLEANING_KEYS = ['clean_name', 'deduplicate', 'validate_identity', 'complete_fields'];
     const DEFAULT_ENRICHMENT_KEYS = ['credit_no', 'legal_rep', 'reg_capital', 'establish_date', 'reg_status'];
-    const DEFAULT_SESSION_PROMPT = '请帮我清洗并补全企业名单。可点击输入框左上角「提示词生成」录入名单、上传 Excel 或图片，也可直接修改本段任务说明后开始。';
+    const DEFAULT_SESSION_PROMPT = '请帮我清洗并补全企业名单。请粘贴企业名称，或上传 Excel / 名单图片，并说明需要补全的字段；也可点击左上角「提示词生成」设置匹配与输出。例如：补全【企业名单】的统一社会信用代码、法定代表人和联系方式，结果导出为 Excel。';
+    const INITIAL_DRAFT_TEMPLATE = Object.freeze({ id: 'dsh-initial-draft/data-cleaning/1', version: 1,
+      fingerprint: 'sha256:2b84b86f669e64e970381e4c7ab8547c77d5e10ee7d5441543c893b9f1272e2b', text: DEFAULT_SESSION_PROMPT });
+    const INITIAL_DRAFT_LEDGER = Symbol.for('dsh.data-cleaning.initial-draft.v1');
+    const compositionState = document[Symbol.for('dsh.data-cleaning.composition.v1')] ||= { active: false };
+    let initialDraftEpoch = 0;
     const CLEANING_SESSION_STORAGE_KEYS = [
       'dsh.data-cleaning-agent.active-session.v2',
       'dsh.data-cleaning-agent.sessions.v1',
@@ -1827,28 +1926,17 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
     }
 
     function isKnownCleaningDraft(value) {
-      if (value === DEFAULT_SESSION_PROMPT) return true;
-      if (typeof value !== 'string') return false;
-      const wizardDraft = value.startsWith('请执行一项企业名单数据清洗补全任务。')
-        && value.includes('企查查连接、套餐额度和费用均由当前用户自己的账号承担。')
-        && value.includes('提供结果和待复核清单的导出。');
-      const executionDraft = value.startsWith('请执行已在「数据清洗补全工作台」确认的企业数据任务。')
-        && value.includes('安全任务凭证：dcq-')
-        && value.includes('data_cleaning_qcc_run');
-      const imageDraft = value.startsWith('请识别我刚刚在向导中安全暂存的企业名单图片')
-        && value.includes('安全图片凭证：dci-')
-        && value.includes('data_cleaning_extract_image_companies');
-      return wizardDraft || executionDraft || imageDraft;
+      // Exact text is the fingerprint: no trimming, substring or fuzzy recognition.
+      return value === INITIAL_DRAFT_TEMPLATE.text;
     }
 
-    function clearCleaningDraft(ctx, sessionId, onlyDefault = false) {
-      if (!sessionId) return false;
-      const conversation = typeof ctx.get === 'function' ? ctx.get('conversation') : ctx.conversation;
-      const shell = conversation?.input?.shell?.(sessionId);
-      if (!shell || typeof shell.setDraft !== 'function') return false;
-      if (onlyDefault && !isKnownCleaningDraft(shell.snapshot?.draft)) return false;
-      shell.setDraft('');
-      return true;
+    function wizardListEntries(value) {
+      if (isKnownCleaningDraft(value)) return [];
+      return String(value || '').split(/\r?\n/).map((line) => line.trim()).filter(Boolean);
+    }
+
+    function invalidListInstruction(values) {
+      return values.some((value) => /【[^】]*】|^(?:请|例如[：:]|输入来源[：:]|结果导出|需要补全)/u.test(value));
     }
 
     function deactivateCleaningSession(sessionId = activeCleaningSessionId) {
@@ -1889,43 +1977,19 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
      */
     function installSessionOwnershipBridge(ctx) {
       const releaseNewSessionBridge = installNewSessionBridge(ctx);
-      let staleDraftObserver = null;
-      let staleDraftObserverTimer = null;
-      const clearStaleDefaultDraft = () => {
-        if (activeCleaningSessionId) return false;
-        const currentSessionId = ctx.sessions?.list?.getSnapshot?.()?.current;
-        const cleared = clearCleaningDraft(ctx, currentSessionId, true);
-        if (cleared && staleDraftObserver) {
-          staleDraftObserver.disconnect();
-          staleDraftObserver = null;
-        }
-        return cleared;
-      };
+      // Observe composition only; never read/write another product's draft.
+      const compositionStart = () => { compositionState.active = true; };
+      const compositionEnd = () => { compositionState.active = false; };
+      document.addEventListener?.('compositionstart', compositionStart, true);
+      document.addEventListener?.('compositionend', compositionEnd, true);
       const leaveSubsystem = () => {
         const sessionId = activeCleaningSessionId;
-        if (sessionId && !ownedSessionId(sessionId)) clearCleaningDraft(ctx, sessionId, true);
         deactivateCleaningSession(sessionId);
       };
       // Only our independently-created native session namespace may regain ownership.
       const initialSession = ctx.sessions?.list?.getSnapshot?.()?.current;
       if (ownedSessionId(initialSession)) markCleaningSession(initialSession);
-      clearStaleDefaultDraft();
-      // DSH 会在插件 apply() 之后异步恢复 composer draft。在首屏稳定期内
-      // 监听 DOM 变化并仅清理完全匹配本插件默认文案，或同时命中
-      // 提示词向导固定首尾与费用声明签名的草稿；不触碰用户自写内容。
-      if (typeof MutationObserver === 'function' && document.documentElement) {
-        staleDraftObserver = new MutationObserver(clearStaleDefaultDraft);
-        staleDraftObserver.observe(document.documentElement, {
-          childList: true,
-          characterData: true,
-          subtree: true,
-        });
-        staleDraftObserverTimer = setTimeout(() => {
-          staleDraftObserver?.disconnect();
-          staleDraftObserver = null;
-          staleDraftObserverTimer = null;
-        }, 10_000);
-      }
+      // Restore ownership only. Refresh/remount/navigation must never seed or clear drafts.
       const unsubscribe = ctx.sessions?.list?.subscribe?.((snapshot) => {
         // Store subscriptions may signal without a payload; read authoritative state.
         const selected = ctx.sessions?.list?.getSnapshot?.() ?? snapshot;
@@ -1936,14 +2000,13 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
         }
         if (!activeCleaningSessionId && ownedSessionId(nextSessionId)) {
           markCleaningSession(nextSessionId);
-        } else if (!activeCleaningSessionId) {
-          clearStaleDefaultDraft();
         }
       });
       return () => {
+        initialDraftEpoch++;
         releaseNewSessionBridge();
-        staleDraftObserver?.disconnect();
-        if (staleDraftObserverTimer !== null) clearTimeout(staleDraftObserverTimer);
+        document.removeEventListener?.('compositionstart', compositionStart, true);
+        document.removeEventListener?.('compositionend', compositionEnd, true);
         if (typeof unsubscribe === 'function') unsubscribe();
       };
     }
@@ -2031,6 +2094,17 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
     }
 
     const MAPPING_ALIASES = {
+      province: ['省', '所属省份'], city: ['市', '所属市', '所属城市'],
+      district: ['区县', '所属区县', '所属区域'], area_code: ['区域编码', '地区编码', '行政区划代码'],
+      industry_section: ['国标行业一级', '国标行业门类名称'],
+      industry_large: ['国标行业二级', '国标行业大类名称'],
+      industry_middle: ['国标行业三级', '国标行业中类名称'],
+      industry_small: ['国标行业四级', '国标行业小类名称'],
+      qcc_industry_level1: ['企查查一级行业', '企查查行业一级分类'],
+      qcc_industry_level2: ['企查查二级行业', '企查查行业二级分类'],
+      qcc_industry_level3: ['企查查三级行业', '企查查行业三级分类'],
+      qcc_industry_level4: ['企查查四级行业', '企查查行业四级分类'],
+      main_products: ['主要产品'], company_scale: ['企业规模分类'],
       beneficial_owner_first_name: ['首条受益所有人名称'],
       actual_controller_name: ['实控人', '实际控制人', '企业实控人名称', '企业实际控制人', '企业实控人名称（自然人请填写姓名）'],
       actual_controller_direct_ratio: ['实际控制人直接持股比例'],
@@ -2406,6 +2480,32 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       return cacheWorkflowTask(actions, sessionId, response.task, reason);
     }
 
+    async function verifyRestoredRuntime(actions, sessionId, task, runtime) {
+      if (!runtime?.restoredFromSession) return runtime;
+      if (!task?.id || !task.source?.checksum || !runtime.sourceRows?.length) {
+        if (task?.id) removeRuntime(task.id);
+        actions.setDataset?.(null);
+        throw new Error('刷新缓存缺少同源校验信息，请重新载入原文件或原文本。');
+      }
+      const response = await requestJson(
+        `/data-cleaning/api/workflow/tasks/${encodeURIComponent(task.id)}/verify-source`,
+        'POST',
+        {
+          originSessionId: sessionId,
+          originWorkspaceId: task.originWorkspaceId,
+          headers: runtime.headers,
+          rows: runtime.sourceRows,
+        },
+      );
+      if (response.verified !== true) {
+        removeRuntime(task.id);
+        actions.setDataset?.(null);
+        throw new Error('刷新缓存与原任务不一致，已停止本次操作。请重新载入原文件或新建任务。');
+      }
+      runtime.restoredFromSession = false;
+      return runtime;
+    }
+
     async function stageConfirmedWorkflow(actions, sessionId, expected = null) {
       const host = await requestJson('/data-cleaning/api/g5/capabilities');
       if (host?.workflowExecutionVersion !== 1) throw new Error('页面与 Host 版本不一致。请保存任务后升级并完整重启 DSH；当前 Host 不支持自动生成结果文件。');
@@ -2427,9 +2527,10 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
           throw new Error('工作台保存的范围与向导不一致，请重试确认；未生成执行凭证。');
         }
       }
-      const runtime = runtimeFor(task.id, false);
+      const runtime = runtimeFor(task.id, false, task.source?.checksum || '');
       if (!task.fieldSelection?.length) throw new Error('当前未选择企查查补全字段。请在规则页选择补全字段，或使用工作台本地清洗与结果导出；不会发起企查查调用。');
       if (!runtime?.rows?.length) throw new Error('Host 名单尚未载入，请先解析数据。');
+      await verifyRestoredRuntime(actions, sessionId, task, runtime);
       if (task.state === 'uploaded') task = await workflowAction(actions, sessionId, task, 'rules', {
         objectives: task.objectives, fieldSelection: task.fieldSelection, mappings: task.mappings, matchRules: task.matchRules,
       });
@@ -2725,13 +2826,45 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       return `${CLEANING_SESSION_ID_PREFIX}${globalThis.crypto.randomUUID()}`;
     }
 
+    function emptyInitialInput(snapshot) {
+      // Missing Host capabilities fail closed. draftRev also detects type-then-clear.
+      return snapshot?.phase === 'plain' && snapshot.draft === ''
+        && snapshot.draftRev === 0 && Array.isArray(snapshot.imageIds) && snapshot.imageIds.length === 0
+        && Array.isArray(snapshot.occurrences) && snapshot.occurrences.length === 0;
+    }
+
+    async function initializeCleaningDraft(ctx, sessionId, entry) {
+      const ledger = document[INITIAL_DRAFT_LEDGER] ||= new Map();
+      if (!ownedSessionId(sessionId) || entry?.createdId !== sessionId || ledger.has(sessionId)) return false;
+      // Reserve before yielding: no duplicate initialization, even on plugin remount.
+      ledger.set(sessionId, { templateId: INITIAL_DRAFT_TEMPLATE.id, state: 'skipped' });
+      const conversation = typeof ctx.get === 'function' ? ctx.get('conversation') : ctx.conversation;
+      const shell = conversation?.input?.shell?.(sessionId);
+      const eligible = () => !compositionState.active && !entry.interrupted()
+        && ctx.sessions.list.getSnapshot().current === entry.previousSession
+        && typeof shell?.setDraft === 'function' && typeof shell.state?.subscribe === 'function'
+        && emptyInitialInput(shell.snapshot);
+      if (!eligible()) return false;
+      const revision = shell.snapshot.draftRev;
+      let changed = false;
+      const release = shell.state?.subscribe?.(() => { if (!emptyInitialInput(shell.snapshot)) changed = true; });
+      try {
+        await Promise.resolve();
+        // Authoritative second snapshot, not captured render props. No await between CAS and write.
+        if (changed || !eligible() || shell.snapshot.draftRev !== revision) return false;
+        shell.setDraft(INITIAL_DRAFT_TEMPLATE.text);
+        const accepted = shell.snapshot?.draft === INITIAL_DRAFT_TEMPLATE.text;
+        ledger.set(sessionId, { templateId: INITIAL_DRAFT_TEMPLATE.id, state: accepted ? 'initialized' : 'skipped' });
+        return accepted;
+      } finally { if (typeof release === 'function') release(); }
+    }
+
     /** 使用 DSH 原生工作区、会话和输入机打开中央对话，不构造第二套聊天界面。 */
     async function startCleaningSession(ctx) {
       const workspace = ctx.workspaces.list.getSnapshot();
       const current = ctx.sessions.list.getSnapshot().current;
       if (ownedSessionId(current)) {
         markCleaningSession(current);
-        requestWorkbenchOpen(workflowTaskBySession.get(String(current))?.stage || 'upload', current);
         return current;
       }
       const items = Array.isArray(workspace.items) ? workspace.items : [];
@@ -2752,16 +2885,34 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
         throw new Error('当前 DSH 版本没有可用的独立会话创建能力');
       }
       const requestedId = createCleaningSessionId();
-      const sessionId = await ctx.sessions.create({ workspaceId: targetWorkspace.workspaceId, sessionId: requestedId });
-      if (sessionId !== requestedId) {
-        throw new Error('当前 DSH 版本不支持独立会话创建，请升级 DSH 后再打开数据清洗补全智能体');
+      const entryEpoch = initialDraftEpoch;
+      let interrupted = compositionState.active;
+      const interrupt = () => { interrupted = true; };
+      const events = ['beforeinput', 'input', 'paste', 'drop', 'compositionstart'];
+      for (const event of events) document.addEventListener?.(event, interrupt, true);
+      const stopNavigation = ctx.sessions.list.subscribe?.(() => {
+        if (ctx.sessions.list.getSnapshot().current !== current) interrupt();
+      });
+      let sessionId;
+      try {
+        sessionId = await ctx.sessions.create({ workspaceId: targetWorkspace.workspaceId, sessionId: requestedId });
+        if (sessionId !== requestedId) {
+          throw new Error('当前 DSH 版本不支持独立会话创建，请升级 DSH 后再打开数据清洗补全智能体');
+        }
+        await initializeCleaningDraft(ctx, sessionId, {
+          createdId: requestedId, previousSession: current,
+          interrupted: () => interrupted || entryEpoch !== initialDraftEpoch,
+        });
+        // A late create must not bring the user back from another Session/product.
+        if (!interrupted && entryEpoch === initialDraftEpoch && ctx.sessions.list.getSnapshot().current === current) {
+          ctx.sessions.open(sessionId);
+          markCleaningSession(sessionId);
+        }
+        return sessionId;
+      } finally {
+        if (typeof stopNavigation === 'function') stopNavigation();
+        for (const event of events) document.removeEventListener?.(event, interrupt, true);
       }
-      const conversation = typeof ctx.get === 'function' ? ctx.get('conversation') : ctx.conversation;
-      if (!conversation) throw new Error('DSH 对话服务尚未就绪，请稍后重试');
-      conversation.input.shell(sessionId).setDraft(DEFAULT_SESSION_PROMPT);
-      ctx.sessions.open(sessionId);
-      markCleaningSession(sessionId);
-      return sessionId;
     }
 
     /**
@@ -3385,6 +3536,10 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       return rows.map((row) => {
         const name = nameField ? String(row?.[nameField] ?? '').trim() : '';
         const credit = creditField ? String(row?.[creditField] ?? '').trim() : '';
+        if (isKnownCleaningDraft(name)) return '';
+        if (invalidListInstruction([name, credit])) {
+          throw new Error('名单包含占位符、示例或任务说明，请提供实际企业名称或信用代码后再继续。');
+        }
         return [name, credit].filter(Boolean).join(' | ');
       }).filter(Boolean).slice(0, 50);
     }
@@ -3465,6 +3620,10 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       });
       const [busy, setBusy] = react.useState(false);
       const [error, setError] = react.useState(null);
+      const [draftPolicy, setDraftPolicy] = react.useState('cancel');
+      const lastWizardDraft = react.useRef(null);
+
+      react.useEffect(() => { setDraftPolicy('cancel'); lastWizardDraft.current = null; }, [sessionId]);
 
       react.useEffect(() => {
         const commandId = imageCommand?.commandId;
@@ -3625,11 +3784,12 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
         }
       };
       const selectedEntries = () => mode === 'text'
-        ? rawText.split(/\r?\n/).map((line) => line.trim()).filter(Boolean)
+        ? wizardListEntries(rawText)
         : entries;
       const validateSource = () => {
         const values = selectedEntries();
         if (mode === 'text' && !values.length) return '请先录入企业名称或统一社会信用代码，每行一条。';
+        if (mode === 'text' && invalidListInstruction(values)) return '请提供实际企业名单，每行一个名称或信用代码；请勿将【企业名单】、示例或任务说明作为企业。';
         if (mode === 'text' && values.some(isImagePathEntry)) return '检测到图片文件路径，不是企业名单。请切换「上传图片识别」，重新选择或粘贴图片。';
         if (mode === 'excel' && !values.length) return '请先上传并解析 Excel / CSV / JSON 文件。';
         if (mode === 'image' && !imageCommand) return '请先选择、粘贴或拖入一张包含企业名单的图片。';
@@ -3661,16 +3821,26 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
           return;
         }
         const inputSnapshot = getInputSnapshot?.(sessionId);
-        if (['adjudicating', 'submitting'].includes(inputSnapshot?.phase)) {
+        if (!inputSnapshot || inputSnapshot.phase !== 'plain' || compositionState.active) {
           setError('当前对话正在提交，请等待完成后再回填。图片与字段选择已保留。');
           return;
         }
+        const unchangedWizard = lastWizardDraft.current?.sessionId === sessionId
+          && lastWizardDraft.current.draft === inputSnapshot.draft;
+        const userDraft = inputSnapshot.draft && !isKnownCleaningDraft(inputSnapshot.draft) && !unchangedWizard;
+        if (userDraft && draftPolicy === 'cancel') {
+          setError('已保留当前草稿。请选择“替换”或“追加”后回填，或关闭向导取消。'); return;
+        }
+        const prefix = unchangedWizard ? lastWizardDraft.current.prefix
+          : userDraft && draftPolicy === 'append' ? inputSnapshot.draft + '\n\n' : '';
         const config = promptConfig();
         if (mode === 'image' && imageCommand?.state === 'prepared' && imageAttachment && typeof removeImage === 'function') {
           // qcc-document-mcp 只读取 Host 暂存文件；发送前释放原生附件，文本模型不会收到图片。
           removeImage(sessionId, imageAttachment);
           setImageAttachment(null);
         }
+        const attachmentSnapshot = JSON.stringify({imageIds:getInputSnapshot?.(sessionId)?.imageIds,
+          occurrences:getInputSnapshot?.(sessionId)?.occurrences});
         setBusy(true);
         try {
           if (mode === 'text') {
@@ -3697,9 +3867,18 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
             prompt = mode === 'image' && imageCommand?.state === 'prepared' ? command.prompt : command.prompt + '\n\n' + buildTaskPrompt(config);
             if (mode === 'image' && imageCommand?.state === 'prepared') setImageCommand(command);
           }
+          const latest = getInputSnapshot?.(sessionId);
+          if (!isCleaningSession(sessionId) || compositionState.active || latest?.phase !== 'plain'
+            || latest.draft !== inputSnapshot.draft || latest.draftRev !== inputSnapshot.draftRev
+            || JSON.stringify({imageIds:latest.imageIds,occurrences:latest.occurrences}) !== attachmentSnapshot) {
+            throw new Error('草稿或会话已变化，未覆盖您的内容；请确认后再次回填。');
+          }
+          prompt = prefix + prompt;
           inputActions.setDraft(prompt);
           const updatedInput = getInputSnapshot?.(sessionId);
           if (updatedInput && updatedInput.draft !== prompt) throw new Error('当前对话尚未接收任务说明，请稍后重试回填。');
+          lastWizardDraft.current = {sessionId,draft:prompt,prefix};
+          setDraftPolicy('cancel');
         } catch (draftError) {
           setError(draftError instanceof Error ? draftError.message : String(draftError));
           return;
@@ -3963,6 +4142,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
             ),
             h('details', { className: 'dcAgentExtraFields', open: mode !== 'excel' },
             h('summary', null, `可选：追加补全字段（共 ${ENRICHMENT_OPTIONS.length} 项）`),
+            h('p', { className: 'dcAgentHint' }, CORE_FIELD_GUIDE),
             h('label', { className: 'dcAgentPromptField' },
               h('span', null, `查找补全字段 · 已选 ${enrichmentKeys.length}`),
               h('input', { type: 'search', className: 'dcAgentField', value: fieldQuery, 'aria-label': '查找补全字段',
@@ -3990,6 +4170,11 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
           wizardStep === 4 ? h('div', { className: 'dcAgentWizardPane' },
             h('h4', null, '确认任务描述'),
             h('p', null, '回填后仍可在对话框继续编辑；回填本身不会调用企查查 MCP。'),
+            h('label', null, '已有草稿时的回填方式 ',
+              h('select', { 'aria-label':'已有草稿时的回填方式', value:draftPolicy, onChange:event=>setDraftPolicy(event.target.value) },
+                h('option', {value:'cancel'}, '保留草稿（取消回填）'),
+                h('option', {value:'replace'}, '替换已有草稿'),
+                h('option', {value:'append'}, '追加到已有草稿'))),
             h('pre', { className: 'dcAgentPromptPreview' }, buildTaskPrompt(promptConfig())),
           ) : null,
           h('p', { className: 'dcAgentPromptNote' }, '向导只生成任务描述并保存任务设置。只有你在任务中明确确认后，才会使用当前客户自己的企查查 MCP 连接、账号和额度。'),
@@ -4185,6 +4370,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
         sizeBytes: source.sizeBytes || 0,
         checksum: result.checksum || source.checksum || '',
       };
+      runtime.restoredFromSession = false;
       const mappings = guessMappings(runtime.headers);
       const guessedNameField = mappings.find((mapping) => mapping.targetField === 'company_name')?.sourceField
         ?? mappings.find((mapping) => mapping.targetField === 'credit_no')?.sourceField
@@ -4214,6 +4400,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
         sizeBytes: source.sizeBytes || task.source?.sizeBytes || 0,
         checksum: result.checksum || '',
       };
+      runtime.restoredFromSession = false;
       actions.setDataset({
         fmt: result.fmt ?? task.source?.type ?? 'csv',
         headers: runtime.headers,
@@ -4222,6 +4409,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       });
       const primary = (task.mappings || []).find((mapping) => ['company_name', 'credit_no', 'reg_no'].includes(mapping.targetField));
       if (primary) actions.setNameField(primary.sourceField);
+      persistRuntimeSession(task.id, runtime);
       return runtime;
     }
 
@@ -4244,6 +4432,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
           sizeBytes: source.sizeBytes || 0,
           checksum: result.checksum || source.checksum || '',
       };
+      target.restoredFromSession = false;
       if (['draft', 'uploaded', 'parse_failed'].includes(current.state)) {
         current = await workflowAction(actions, sessionId, current, 'upload', {
           source: {
@@ -4258,6 +4447,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
         }, 'import');
       }
       if (current.state === 'uploaded') current = await updateWorkflowTask(actions, sessionId, current, importedDraft);
+      persistRuntimeSession(current.id, target, stagingKey);
       return current;
     }
 
@@ -4443,7 +4633,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       const objectives = useStore((state) => state.objectives);
       const fieldSelection = useStore((state) => state.fieldSelection);
       const matchRules = useStore((state) => state.matchRules);
-      const dataset = useStore((state) => state.dataset);
+      const storedDataset = useStore((state) => state.dataset);
       const profile = useStore((state) => state.profile);
       const clean = useStore((state) => state.clean);
       const complete = useStore((state) => state.complete);
@@ -4455,11 +4645,11 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       const paidConfirmed = useStore((state) => state.paidConfirmed);
       const activeSessionId = useStore((state) => state.activeSessionId);
       const [fieldQuery, setFieldQuery] = react.useState('');
-      const [intakeEditorOpen, setIntakeEditorOpen] = react.useState(() => dataset == null);
+      const [intakeEditorOpen, setIntakeEditorOpen] = react.useState(() => storedDataset == null);
 
       react.useEffect(() => {
-        setIntakeEditorOpen(dataset == null);
-      }, [dataset]);
+        setIntakeEditorOpen(storedDataset == null);
+      }, [storedDataset]);
 
       react.useEffect(() => {
         if (!open || !visible) return;
@@ -4514,13 +4704,19 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
 
       // Host 决定 Tab 挂载/可见性；内容不再按私有 open flag 隐藏。
 
-      const hasData = dataset !== null && dataset.rowCount > 0;
-      const requiresQcc = objectives.includes('validate_identity') || objectives.includes('complete_fields');
       const cachedTask = workflowTaskBySession.get(String(activeSessionId || 'unassigned')) ?? workflowTask;
       const runtimeKey = cachedTask?.id ?? `session:${activeSessionId || 'unassigned'}`;
-      const runtime = runtimeFor(runtimeKey);
+      const runtime = runtimeFor(runtimeKey, true, cachedTask?.source?.checksum || '');
       const stagedSource = runtimeFor(`session:${activeSessionId || 'unassigned'}`, false);
       const sourceRows = stagedSource?.sourceRows ?? runtime.sourceRows ?? runtime.rows;
+      const dataset = storedDataset ?? (sourceRows.length ? {
+        fmt: runtime.source?.type || cachedTask?.source?.type || 'csv',
+        headers: runtime.headers.length ? runtime.headers : cachedTask?.source?.headers || [],
+        rowCount: cachedTask?.source?.rowCount ?? sourceRows.length,
+        preview: sourceRows.slice(0, 5),
+      } : null);
+      const hasData = dataset !== null && dataset.rowCount > 0;
+      const requiresQcc = objectives.includes('validate_identity') || objectives.includes('complete_fields');
       const intake = intakeState(dataset, sourceRows, input, runtime.source || stagedSource?.source || cachedTask?.source);
       const normalizedStage = normalizeViewStage(step);
       const flowCompatible = cachedTask?.flow?.flowVersion === SUPPORTED_FLOW_VERSION;
@@ -4542,8 +4738,8 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       const showIntakeEditor = canAction('import-data') && (!dataset || intakeEditorOpen);
       const runtimeRecoveryNeeded = Boolean(cachedTask?.source?.rowCount) && runtime.rows.length === 0;
       const runtimeRecoveryAllowed = runtimeRecoveryNeeded
-        && isCurrentView
-        && ['uploaded', 'rules_confirmed', 'diagnosed'].includes(cachedTask?.state)
+        && ((isCurrentView && ['uploaded', 'rules_confirmed', 'diagnosed'].includes(cachedTask?.state))
+          || (normalizedStage === 'upload' && viewAccess === 'read'))
         && cachedTask?.source?.type !== 'image';
       const lastCsv = runtime.lastCsv;
       const fieldByPattern = (pattern) => runtime.headers.find((field) => pattern.test(field)) ?? null;
@@ -4665,6 +4861,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       };
 
       const performProfile = async (task) => {
+        await verifyRestoredRuntime(actions, activeSessionId, task, runtime);
         const r = await api('/data-cleaning/api/mvp/profile', {
           rows: runtime.rows,
           headers: runtime.headers,
@@ -4700,6 +4897,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
         actions.setBusy(true);
         actions.setError(null);
         try {
+          await verifyRestoredRuntime(actions, activeSessionId, cachedTask, runtime);
           const r = await api('/data-cleaning/api/mvp/clean', {
             rows: runtime.rows,
             headers: runtime.headers,
@@ -4728,6 +4926,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
         actions.setBusy(true);
         actions.setError(null);
         try {
+          await verifyRestoredRuntime(actions, activeSessionId, cachedTask, runtime);
           const r = await api('/data-cleaning/api/mvp/complete', {
             rows: runtime.rows,
             headers: runtime.headers,
@@ -4900,6 +5099,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
         actions.setBusy(true);
         actions.setError(null);
         try {
+          await verifyRestoredRuntime(actions, activeSessionId, cachedTask, runtime);
           let current = await ensureWorkflowTask(actions, activeSessionId, {
             title: taskTitle, objectives, fieldSelection, mappings, matchRules,
           });
@@ -4957,9 +5157,9 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
           actions.setError('当前会话的任务向导尚未就绪或正在解析图片，请稍后重试。');
           return;
         }
-        if (cachedTask?.id) runtimeTasks.delete(String(cachedTask.id));
-        runtimeTasks.delete(`session:${activeSessionId || 'unassigned'}`);
-        runtimeTasks.delete('unassigned');
+        if (cachedTask?.id) removeRuntime(cachedTask.id);
+        removeRuntime(`session:${activeSessionId || 'unassigned'}`);
+        removeRuntime('unassigned');
         workflowTaskBySession.delete(String(activeSessionId || 'unassigned'));
         actions.resetWorkflow();
         actions.close();
@@ -4977,6 +5177,7 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
         if (Array.isArray(latest.artifacts) && latest.artifacts.length) return latest;
         const rows = exportRows();
         if (!rows?.length) throw new Error('当前页面没有可导出的明细。若任务来自历史记录，请重新上传源文件并继续处理。');
+        await verifyRestoredRuntime(actions, activeSessionId, latest, runtime);
         const response = await requestJson(`/data-cleaning/api/workflow/tasks/${encodeURIComponent(latest.id)}/artifacts`, 'POST', {
           originSessionId: activeSessionId, originWorkspaceId: latest.originWorkspaceId,
           expectedRevision: latest.revision,
@@ -5256,10 +5457,10 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
             h('details', { className: 'dcAgentExtraFields' },
             h('summary', null, `可选：追加补全字段（共 ${targetFields.filter(([id]) => id !== 'phone').length} 项）`),
             h('div', { className: 'dcAgentRow', 'aria-label': '补全字段批次概览' },
-              h('span', { className: 'dcAgentChip' }, '基础字段 30'),
-              h('span', { className: 'dcAgentChip' }, '第一批 40'),
-              h('span', { className: 'dcAgentChip' }, '第二批 58'),
+              h('span', { className: 'dcAgentChip' }, `工商与简介 ${uiFieldGroups.filter(([id]) => ['company_registration', 'company_profile'].includes(id)).reduce((sum, [, , fields]) => sum + fields.length, 0)} 项`),
+              h('span', { className: 'dcAgentChip' }, `共 ${targetFields.filter(([id]) => id !== 'phone').length} 项可选字段`),
             ),
+            h('p', { className: 'dcAgentHint' }, CORE_FIELD_GUIDE),
             h('label', { className: 'dcAgentPromptField' }, h('span', null, '查找补全字段'),
               h('input', { type: 'search', className: 'dcAgentField', value: fieldQuery, 'aria-label': '查找补全字段',
                 placeholder: '搜索维度、字段或工具名称', onChange: (event) => setFieldQuery(event.target.value) })),
@@ -5287,22 +5488,30 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
           ),
         );
       } else if (step === 'profile') {
+        const persistedQuality = cachedTask?.qualitySummary;
         pane = h('div', { className: 'dcAgentPane dcAgentProfilePane' },
           h('section', { className: 'dcAgentSection dcAgentProfileSurface', 'aria-label': '质量体检报告' },
             h('header', { className: 'dcAgentProfileHeader' },
               h('h3', null, '质量体检报告'),
               h('p', { className: 'dcAgentHint' }, '基于已解析数据生成本地质量画像（缺失率、去重值、金额分布）。本步骤不发起企查查调用。'),
             ),
-            profile ? h('div', { className: 'dcAgentProfileReport' },
+            profile || persistedQuality ? h('div', { className: 'dcAgentProfileReport' },
+              !profile ? h('p', { className: 'dcAgentHint', role: 'status' }, '以下为 Host 保存的体检摘要；字段级明细未持久化。') : null,
               h('div', { className: 'dcAgentGrid dcAgentProfileMetrics', 'aria-label': '体检摘要' },
-                stat('行数', profile.rowCount ?? '—'),
-                stat('列数', profile.columnCount ?? '—'),
-                profile.amountStats ? stat('金额 最小', profile.amountStats.min) : null,
-                profile.amountStats ? stat('金额 最大', profile.amountStats.max) : null,
-                profile.amountStats ? stat('金额 总和', profile.amountStats.sum) : null,
-                profile.amountStats ? stat('金额 均值', typeof profile.amountStats.mean === 'number' ? profile.amountStats.mean.toFixed(2) : profile.amountStats.mean) : null,
+                stat('行数', profile?.rowCount ?? persistedQuality?.total ?? cachedTask?.source?.rowCount ?? '—'),
+                stat('列数', profile?.columnCount ?? cachedTask?.source?.columnCount ?? '—'),
+                profile?.amountStats ? stat('金额 最小', profile.amountStats.min) : null,
+                profile?.amountStats ? stat('金额 最大', profile.amountStats.max) : null,
+                profile?.amountStats ? stat('金额 总和', profile.amountStats.sum) : null,
+                profile?.amountStats ? stat('金额 均值', typeof profile.amountStats.mean === 'number' ? profile.amountStats.mean.toFixed(2) : profile.amountStats.mean) : null,
+                !profile && persistedQuality ? stat('有效主体', persistedQuality.valid, persistedQuality.valid > 0 ? 'good' : null) : null,
+                !profile && persistedQuality ? stat('缺少主体标识', persistedQuality.missingAnchor, persistedQuality.missingAnchor > 0 ? 'bad' : null) : null,
+                !profile && persistedQuality ? stat('重复主体', persistedQuality.duplicates, persistedQuality.duplicates > 0 ? 'warn' : null) : null,
+                !profile && persistedQuality ? stat('信用代码异常', persistedQuality.invalidCreditNo, persistedQuality.invalidCreditNo > 0 ? 'warn' : null) : null,
+                !profile && persistedQuality ? stat('联系电话异常', persistedQuality.invalidPhone, persistedQuality.invalidPhone > 0 ? 'warn' : null) : null,
+                !profile && persistedQuality ? stat('空白字段', persistedQuality.emptyFields, persistedQuality.emptyFields > 0 ? 'warn' : null) : null,
               ),
-              Array.isArray(profile.columns) && profile.columns.length ? h('section', { className: 'dcAgentProfileTableSection', 'aria-label': '字段质量明细' },
+              Array.isArray(profile?.columns) && profile.columns.length ? h('section', { className: 'dcAgentProfileTableSection', 'aria-label': '字段质量明细' },
                 h('div', { className: 'dcAgentProfileTableHead' },
                   h('h4', null, '字段质量明细'),
                   h('span', { className: 'dcAgentHint' }, `共 ${profile.columns.length} 列`),
@@ -5393,9 +5602,22 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
                 canAction('resolve-candidate') ? h('button', { type: 'button', className: 'dcAgentButton', disabled: busy || !paidConfirmed, onClick: () => resolveCandidate(item, candidate) }, '确认此主体') : null,
               )),
             )),
+          ) : cachedTask?.matchSummary ? h('section', { className: 'dcAgentSection', 'aria-label': '主体匹配结果摘要' },
+            h('h3', null, '主体匹配结果'),
+            h('p', { className: 'dcAgentHint', role: 'status' }, '以下为 Host 保存的任务摘要；逐行结果和候选详情未持久化。'),
+            h('div', { className: 'dcAgentGrid' },
+              stat('总数', cachedTask.matchSummary.total),
+              stat('精确匹配', (cachedTask.matchSummary.exact ?? 0) + (cachedTask.matchSummary.confirmed ?? 0), 'good'),
+              stat('候选记录', cachedTask.matchSummary.candidate, cachedTask.matchSummary.candidate > 0 ? 'warn' : null),
+              stat('待核验', cachedTask.matchSummary.reviewRequired, cachedTask.matchSummary.reviewRequired > 0 ? 'warn' : null),
+              stat('未匹配', cachedTask.matchSummary.unresolved, cachedTask.matchSummary.unresolved > 0 ? 'warn' : null),
+              stat('失败', cachedTask.matchSummary.failed, cachedTask.matchSummary.failed > 0 ? 'bad' : null),
+            ),
           ) : null,
         );
       } else if (step === 'enrich') {
+        const persistedEnrichment = cachedTask?.enrichmentSummary;
+        const hasRuntimeEnrichment = Boolean(clean || complete || qccRun);
         pane = h('div', { className: 'dcAgentPane' },
           h('p', { className: 'dcAgentHint' }, '执行本地确定性清洗并核对本任务所选补全字段。外部字段以当前用户企查查 MCP 的真实返回为准，缺失值不会被编造。'),
           h('section', { className: 'dcAgentSection' },
@@ -5414,13 +5636,23 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
               stat('重复', clean.badDuplicate, clean.badDuplicate > 0 ? 'warn' : null),
             ) : null,
           ),
-          h('div', { className: 'dcAgentGrid' },
-            stat('输入行数', dataset ? dataset.rowCount : '—'),
-            stat('清洗保留', clean ? clean.kept : '—', clean && clean.kept > 0 ? 'good' : null),
-            stat('本地补全', complete ? complete.completed : '—', complete && complete.completed > 0 ? 'good' : null),
-            stat('QCC 已补全', qccEnrichedCount, qccEnrichedCount > 0 ? 'good' : null),
-            stat('待核验', qccReviewCount, qccReviewCount > 0 ? 'warn' : null),
-          ),
+          persistedEnrichment && !hasRuntimeEnrichment ? h('section', { className: 'dcAgentSection', 'aria-label': '字段补全结果摘要' },
+            h('h3', null, '字段补全结果'),
+            h('p', { className: 'dcAgentHint', role: 'status' }, '以下为 Host 保存的任务摘要；逐行补全结果请从结果下载查看。'),
+            h('div', { className: 'dcAgentGrid' },
+              stat('总数', persistedEnrichment.total),
+              stat('已补全', persistedEnrichment.completed, persistedEnrichment.completed > 0 ? 'good' : null),
+              stat('原值保留', persistedEnrichment.unchanged),
+              stat('待核验', persistedEnrichment.reviewRequired, persistedEnrichment.reviewRequired > 0 ? 'warn' : null),
+              stat('失败', persistedEnrichment.failed, persistedEnrichment.failed > 0 ? 'bad' : null),
+            ),
+          ) : h('div', { className: 'dcAgentGrid' },
+              stat('输入行数', dataset ? dataset.rowCount : '—'),
+              stat('清洗保留', clean ? clean.kept : '—', clean && clean.kept > 0 ? 'good' : null),
+              stat('本地补全', complete ? complete.completed : '—', complete && complete.completed > 0 ? 'good' : null),
+              stat('QCC 已补全', qccEnrichedCount, qccEnrichedCount > 0 ? 'good' : null),
+              stat('待核验', qccReviewCount, qccReviewCount > 0 ? 'warn' : null),
+            ),
           canAction('create-local-artifacts') ? h('div', { className: 'dcAgentRow' }, h('button', {
             type: 'button', className: 'dcAgentButton is-primary', disabled: busy || !exportRows()?.length,
             onClick: () => createAndDownload('complete', 'xlsx'),
@@ -6114,7 +6346,6 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       DatasetReview,
       reviewColumns,
       reviewCsv,
-      clearCleaningDraft,
       entriesToDataset,
       extractPromptEntries,
       imageFilesFromTransfer,
@@ -6123,6 +6354,8 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       guessMappings,
       intakeState,
       restoreParsedRuntime,
+      runtimeFor,
+      verifyRestoredRuntime,
       projectCompletionResult,
       mappingRecommendations,
       mappedOutputFields,
@@ -6135,6 +6368,12 @@ body:has(.dcAgentPromptBackdrop) { overflow: hidden; }
       installNewSessionBridge,
       isCleaningSession,
       isKnownCleaningDraft,
+      wizardListEntries,
+      invalidListInstruction,
+      INITIAL_DRAFT_TEMPLATE,
+      initializeCleaningDraft,
+      emptyInitialInput,
+      startCleaningSession,
       markCleaningSession,
       qualitySummaryFor,
       plainEntityListDataset,

package/lib/qcc-field-catalog.js

@@ -102,8 +102,16 @@ export const QCC_FIELD_CATALOG = Object.freeze([
     field('registered_address', '注册地址', { defaultSelected: true }),
     field('mailing_address', '通信地址'),
     field('region', '所属地区', { matchAuxiliary: true }),
+    field('province', '省份', { sourcePath: '地区信息.省份' }),
+    field('city', '城市', { sourcePath: '地区信息.城市' }),
+    field('district', '区域', { sourcePath: '地区信息.区域' }),
+    field('area_code', '地区代码', { sourcePath: '地区信息.地区代码' }),
     field('business_scope', '经营范围'),
     field('industry_category', '国标行业'),
+    field('industry_section', '国标行业门类', { sourcePath: '国标行业.门类' }),
+    field('industry_large', '国标行业大类', { sourcePath: '国标行业.大类' }),
+    field('industry_middle', '国标行业中类', { sourcePath: '国标行业.中类' }),
+    field('industry_small', '国标行业小类', { sourcePath: '国标行业.小类' }),
     field('operating_period', '营业期限'),
     field('company_size', '人员规模'),
     field('insured_count', '参保人数'),
@@ -111,6 +119,12 @@ export const QCC_FIELD_CATALOG = Object.freeze([
   ], { releaseBatch: 'current' }),
   group('company_profile', '企业简介', 'get_company_profile', [
     field('qcc_industry', '企查查行业'),
+    field('qcc_industry_level1', '企查查行业一级', { sourcePath: '企查查行业.一级' }),
+    field('qcc_industry_level2', '企查查行业二级', { sourcePath: '企查查行业.二级' }),
+    field('qcc_industry_level3', '企查查行业三级', { sourcePath: '企查查行业.三级' }),
+    field('qcc_industry_level4', '企查查行业四级', { sourcePath: '企查查行业.四级' }),
+    field('main_products', '主营产品', { sourcePath: '主营产品' }),
+    field('company_scale', '企业规模', { sourcePath: '企业规模' }),
     field('company_profile', '企业简介'),
     field('industry_chain_overview', '产业链概览'),
   ], { releaseBatch: 'current' }),

package/lib/qcc.js

@@ -375,8 +375,16 @@ export function mapRegistrationFields(value, fallback = {}) {
     registered_address: text('注册地址', '住所', '经营场所'),
     mailing_address: text('通信地址'),
     region: text('所属地区'),
+    province: coreFieldText(value.地区信息?.省份),
+    city: coreFieldText(value.地区信息?.城市),
+    district: coreFieldText(value.地区信息?.区域),
+    area_code: coreFieldText(value.地区信息?.地区代码),
     business_scope: text('经营范围'),
-    industry_category: text('国标行业'),
+    industry_category: industryDisplay(value.国标行业, ['门类', '大类', '中类', '小类']),
+    industry_section: coreFieldText(value.国标行业?.门类),
+    industry_large: coreFieldText(value.国标行业?.大类),
+    industry_middle: coreFieldText(value.国标行业?.中类),
+    industry_small: coreFieldText(value.国标行业?.小类),
     operating_period: text('营业期限', '经营期限'),
     company_size: text('人员规模'),
     insured_count: text('参保人数'),
@@ -384,16 +392,36 @@ export function mapRegistrationFields(value, fallback = {}) {
   };
 }
 
+// Explicit leaves only: old flat text must not be guessed into industry levels.
+function coreFieldText(value) {
+  return typeof value === 'string' || typeof value === 'number' ? String(value) : '';
+}
+
+function industryDisplay(value, levels) {
+  if (!isRecord(value)) return coreFieldText(value);
+  return levels.map(level => {
+    const text = coreFieldText(value[level]);
+    return text.trim() ? `${level}：${text}` : '';
+  }).filter(Boolean).join('；');
+}
+
 export function mapProfileFields(value) {
   if (!isRecord(value)) {
     throw new QccBridgeError('QCC_CONTRACT_MISMATCH', 'QCC profile tool returned a non-object result');
   }
   if (value.无匹配项 !== undefined) return {};
   return {
-    // 工具只返回“企查查行业”这一最细层级展示值，不得猜测为一级或二级行业。
-    qcc_industry: String(value.企查查行业 ?? ''),
-    company_profile: String(value.企业简介 ?? value.简介 ?? ''),
-    industry_chain_overview: String(value.产业链概览 ?? ''),
+    qcc_industry: industryDisplay(value.企查查行业, ['一级', '二级', '三级', '四级']),
+    qcc_industry_level1: coreFieldText(value.企查查行业?.一级),
+    qcc_industry_level2: coreFieldText(value.企查查行业?.二级),
+    qcc_industry_level3: coreFieldText(value.企查查行业?.三级),
+    qcc_industry_level4: coreFieldText(value.企查查行业?.四级),
+    main_products: Array.isArray(value.主营产品)
+      ? value.主营产品.filter(item => typeof item === 'string' && item.trim()).join('；')
+      : coreFieldText(value.主营产品),
+    company_scale: coreFieldText(value.企业规模),
+    company_profile: coreFieldText(value.企业简介 ?? value.简介),
+    industry_chain_overview: coreFieldText(value.产业链概览),
   };
 }
 

package/lib/skill-enrich.js

@@ -52,6 +52,7 @@ export function registerEnrichSkill(skills) {
       'When the user gives company names and asks to enrich / complete them with Qichacha (企查查), including registration fields, company panorama, ownership penetration, governance, historical changes, risk tags, intellectual property, or operational / bidding dimensions.',
     source: 'dsh-data-cleaning-agent',
     content: [
+      'UX-49 最高优先级：先检查本业务请求的信息是否充分，再考虑以下任何工具或执行规则。若仍含未补全的【】占位符，或没有实际企业名单/Excel/名单图片与待补字段，先用中文要求提供名单或文件及待补字段；零工具调用，包括 OAuth 检查、体检、OCR、MCP、创建 workflow。例子、说明、【企业名单】均不是企业。用户改过的初始模板按普通内容理解，不得整体忽略。有真实暂存来源和已选字段的安全任务凭证可提供这些信息，不要求重复提供。此规则只作用于数据清洗补全业务，不得全局拦截 DSH 发送或其它产品。',
       'You are an enterprise-list enrichment assistant. You fill company lists with Qichacha (QCC) data by calling QCC MCP tools. Never invent, pad, or fabricate any field.',
       '',
       'Workbench execution request (highest priority):',

package/lib/skill.js

@@ -15,6 +15,7 @@ export function registerSkill(skills) {
       'When the user asks to clean, complete, validate, de-duplicate, or summarize a batch of raw data rows, CSV records, or table-like data.',
     source: 'dsh-data-cleaning-agent',
     content: [
+      'UX-49 — Highest priority, before ANY tool: within this data-cleaning business request, if the message still contains an unfilled 【】 placeholder, or lacks actual rows/list/file and requested completion fields, ask the user to provide the actual enterprise list or Excel/list image and the fields to complete. Make ZERO tool calls (including profile, OAuth checks, OCR, MCP and workflow creation). Do not treat examples, instructions or 【企业名单】 as enterprises. Do not intercept other DSH products or the global send action. A valid staged-source task credential can supply the list/field scope; do not demand the same data twice. An edited initial template is ordinary user content, never discard it wholesale.',
       'You are a data cleaning and completion assistant. Work only on the rows the user actually provided; never invent, pad, or fabricate extra rows.',
       '',
       'Workflow:',

package/lib/web.js

@@ -653,6 +653,22 @@ export function mountWebRoutes(wctx, { logger, report, TOOL_NAME, SKILL_NAME })
         return writeJson(res, 405, { ok: false, code: 'DC_METHOD', message: 'Artifact route requires GET or POST.' });
       }
 
+      if (rest.length === 2 && rest[1] === 'verify-source') {
+        if (req.method !== 'POST') {
+          return writeJson(res, 405, { ok: false, code: 'DC_METHOD', message: 'Source verification requires POST.' });
+        }
+        const body = await readBody(req);
+        const payload = body.length ? JSON.parse(body.toString('utf8')) : {};
+        const current = await workflow.assertRequestOrigin(taskId, payload);
+        const verified = Boolean(current.source?.checksum)
+          && sourceVerifier(payload.headers, payload.rows, current.source.checksum) === current.source.checksum;
+        return writeJson(res, 200, {
+          ok: true,
+          marker: 'data-cleaning-source-verifier-v1',
+          verified,
+        });
+      }
+
       if (rest.length !== 3 || rest[1] !== 'actions' || req.method !== 'POST') {
         return writeJson(res, 404, { ok: false, code: 'DC_WORKFLOW_ROUTE', message: 'Workflow route not found.' });
       }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-data-cleaning-agent",
-  "version": "0.9.15",
+  "version": "0.9.17",
   "description": "Data cleaning and data enrichment for CSV/XLSX/JSON enterprise lists in DeepSeek Harness, including spreadsheet cleaning, deduplication, profiling, optional Qichacha MCP and exports.",
   "type": "module",
   "main": "lib/index.js",
@@ -10,6 +10,7 @@
     "./package.json": "./package.json"
   },
   "files": [
+    "docs/CORE-FIELDS-20260918.md",
     "lib",
     "cordis.patch.yml",
     "README.md",