dsh-long-office 0.1.2

package/README.md

@@ -0,0 +1,101 @@
+# dsh-long-office
+
+DSH Web 插件：**Office 读取 + 生成**。读取 Word/Excel/PowerPoint（含旧版格式）为文本/表格，内存解析、不落中间文件；并从 Markdown 生成 Word / Excel / PowerPoint（`md2docx` / `md2xlsx` / `md2pptx`，原 `dsh-long-plugins` 的 `md2docx` 迁入本插件）。
+
+**工具一览**：`office_read`（读 Office）、`md2docx`（Markdown→Word）、`md2xlsx`（Markdown 表格→Excel）、`md2pptx`（Markdown 大纲→PPT，16:9/备注/图片/坐标回显）。
+
+---
+
+## 功能
+
+### 1. 读取 Office → 文本 / 表格（内存解析，模型按需用）
+| 格式 | 依赖 | 说明 |
+| --- | --- | --- |
+| `.docx` | `mammoth` | Word 2007+ → 纯文本 |
+| `.doc` | `word-extractor` | 旧版 Word → 纯文本 |
+| `.xlsx` | `exceljs` | Excel 2007+ → 每工作表表格 |
+| `.xls` | `xlsx` (SheetJS) | 旧版 Excel → 每工作表 CSV |
+| `.pptx` | `jszip` | PPT 2007+ → 每页文本（zip 内 slide XML） |
+| `.ppt` | — | 暂不支持，提示转换为 `.pptx` |
+| `.txt / .md` | — | 直接按 UTF-8 读取 |
+
+读取在内存中完成，**不生成任何中间文件**，模型按需调用 `office_read` 工具查看内容。
+
+### 2. 生成 Office 文档（Markdown → Word / Excel / PowerPoint）
+
+| 工具 | 产物 | 依赖 | 要点 |
+| --- | --- | --- | --- |
+| `md2docx` | `.docx` | `python3` + `python-docx` | 带页码；中文公文排版（宋体＋Times New Roman／1.5 倍行距／首行缩进 2 字符／表格居中／无页眉与装饰线／正文不加粗） |
+| `md2xlsx` | `.xlsx` | `python3` + `openpyxl` | 每个 Markdown 表格一个 sheet（名取上方标题）；表头加粗、细边框、内容居中、**无底纹**、列宽自适应、冻结首行；`=` 开头写成**真公式** |
+| `md2pptx` | `.pptx` | `python3` + `python-pptx` + `Pillow` | 16:9；`# `封面、`## `分页、`- `要点、`> 备注：`演讲者备注、`![说明](图 "宽x高")` 插图；生成后回显每页元素坐标 |
+
+三者默认都**不覆盖**已存在文件（工具的 `overwrite: true` 才覆盖）。脚本路径可分别用配置 `md2docxScript` / `md2xlsxScript` / `md2pptxScript` 覆盖（默认：插件内 `lib/md2docx.py`、`lib/md2xlsx.py`、`lib/md2pptx.py`）。
+
+---
+
+## 安装
+
+在 DSH profile（如 `web`）下执行：
+
+```bash
+# 通过 npm 安装（未发布时用本地 file: 链接）
+dsh plugin --profile web add dsh-long-office
+# 源码调试：
+dsh plugin --profile web add file:./dsh-long-office
+```
+
+`cordis.patch.yml` 会把插件 `dsh-long-office` 插入 profile 的层栈（`inject: [webRuntime]`）。
+
+---
+
+## 提供的服务端能力
+
+- 工具 `office_read`：读取 Office 文件 → 文本/表格。
+- 工具 `md2docx`：Markdown → Word。
+- 路由 `GET /api/dsh-long-office/content?path=<绝对路径>`：返回 Office 文件的内存文本/表格。
+
+> 说明：本插件主要提供服务端工具与路由，`client/client.js` 仅为一个占位模块（无前端 UI）。
+
+---
+
+## 配置
+
+`cordis.patch.yml` 默认复用 DSH Web 内建的 `trustedHosts`，可选覆盖 `md2docxScript`：
+
+```yaml
+- insert:
+    - id: dsh-long-office
+      name: dsh-long-office
+      inject: [webRuntime]
+      config:
+        # 复用内建 Web API 的 Host/Origin 白名单
+        trustedHosts: !!js ctx.webRuntime.trustedHosts
+        # md2docx 脚本覆盖（默认 <plugin>/lib/md2docx.py）
+        # md2docxScript: /your/path/md2docx.py
+```
+
+---
+
+## 验证
+
+安装、重启 DSH 后，验证插件已加载并可用：
+
+```bash
+# 1) 服务已启动
+curl -s -o /dev/null -w "%{http_code}\n" http://127.0.0.1:3080
+# 期望：200
+
+# 2) 读取一个 docx
+curl -s "http://127.0.0.1:3080/api/dsh-long-office/content?path=/absolute/path/to/example.docx"
+# 期望：{"ok":true,"content":"..."}
+
+# 3) 工具可用（通过会话问模型）
+# "用 office_read 读一下 /path/to/example.xlsx"
+# "用 md2docx 把 /path/to/report.md 生成成 Word"
+```
+
+---
+
+## License
+
+MIT

package/client/client.js

@@ -0,0 +1,14 @@
+window.__ModuleLoader__.load({
+  id: 'dsh-long-office',
+  factory: (require) => {
+    const module = { exports: {} }
+    const exports = module.exports
+    Object.defineProperty(exports, Symbol.toStringTag, { value: 'Module' })
+    // dsh-long-office 主要提供服务端 Office 读取/生成工具与路由，此客户端仅为占位(无前端 UI)。
+    function apply() {}
+    const inject = []
+    exports.apply = apply
+    exports.inject = inject
+    return module.exports
+  }
+})

package/cordis.patch.yml

@@ -0,0 +1,10 @@
+# dsh bundle patch: inserts the office plugin into a profile's layer stack.
+- insert:
+    - id: dsh-long-office
+      name: dsh-long-office
+      inject: [webRuntime]
+      config:
+        # Reuse the same trusted Host/Origin authorities as the built-in Web API.
+        trustedHosts: !!js ctx.webRuntime.trustedHosts
+        # md2docx script override (default: <plugin>/lib/md2docx.py).
+        # md2docxScript: /your/path/md2docx.py

package/dsh.plugin.json

@@ -0,0 +1,12 @@
+{
+  "name": "dsh-long-office",
+  "description": "DSH Office read + generate: read Word/Excel/PowerPoint into text/tables in memory (no intermediates, model on-demand); generate styled Word from Markdown (md2docx). | DSH Office 读取 + 生成：内存读取 Office 内容为文本/表格(无中间文件、模型按需用)；Markdown 生成带页码 Word。",
+  "version": "0.1.2",
+  "entry": {
+    "module": "./lib/index.js",
+    "client": "./client/client.js"
+  },
+  "dependencies": [
+    "@deepseek-ai/dsh-tools"
+  ]
+}

package/lib/index.js

@@ -0,0 +1,236 @@
+import { fileURLToPath } from "node:url";
+import { dirname, extname, join, resolve } from "node:path";
+import { readFile } from "node:fs/promises";
+import { spawn } from "node:child_process";
+import mammoth from "mammoth";
+import ExcelJS from "exceljs";
+import XLSX from "xlsx";
+import WordExtractor from "word-extractor";
+import { defineTool } from "@deepseek-ai/dsh-tools";
+
+const PACKAGE_DIR = dirname(fileURLToPath(import.meta.url));
+const MD2DOCX_SCRIPT = resolve(process.env.MD2DOCX_SCRIPT || join(PACKAGE_DIR, "md2docx.py"));
+  const MD2XLSX_SCRIPT = resolve(process.env.MD2XLSX_SCRIPT || join(PACKAGE_DIR, "md2xlsx.py"));
+  const MD2PPTX_SCRIPT = resolve(process.env.MD2PPTX_SCRIPT || join(PACKAGE_DIR, "md2pptx.py"));
+
+function sendJson(res, status, obj) {
+  res.writeHead(status, { "content-type": "application/json; charset=utf-8", "cache-control": "no-store" });
+  res.end(JSON.stringify(obj));
+}
+
+/** 内存读取 docx → 纯文本（mammoth）。 */
+async function docxToText(buf) {
+  const { value } = await mammoth.extractRawText({ buffer: buf });
+  return value;
+}
+
+/** 内存读取 doc → 纯文本（word-extractor）。 */
+async function docToText(buf) {
+  const w = new WordExtractor();
+  const d = await w.extract(Buffer.from(buf));
+  return d.getBody();
+}
+
+/** 内存读取 xlsx → 表格文本（exceljs）。 */
+async function xlsxToText(buf) {
+  const wb = new ExcelJS.Workbook();
+  await wb.xlsx.load(buf);
+  const rows = [];
+  wb.eachSheet((ws) => {
+    rows.push(`## 工作表：${ws.name}`);
+    ws.eachRow((row) => {
+      rows.push((row.values || []).slice(1).map((c) => c == null ? "" : String(c)).join("\t"));
+    });
+    rows.push("");
+  });
+  return rows.join("\n");
+}
+
+/** 内存读取 xls → 表格文本（SheetJS/xlsx，支持 .xls 与 .xlsx）。 */
+function xlsToText(buf) {
+  const wb = XLSX.read(buf, { type: "buffer" });
+  const out = [];
+  for (const name of wb.SheetNames) {
+    out.push(`## 工作表：${name}`);
+    const csv = XLSX.utils.sheet_to_csv(wb.Sheets[name]);
+    out.push(csv);
+    out.push("");
+  }
+  return out.join("\n");
+}
+
+/** 内存读取 pptx → 每页文本（pptx 是 zip + slide XML）。 */
+async function pptxToText(buf) {
+  const jszip = (await import("jszip")).default;
+  const zip = await jszip.loadAsync(buf);
+  const entryNames = Object.keys(zip.files).filter((n) => /^ppt\/slides\/slide\d+\.xml$/i.test(n)).sort((a, b) => a.localeCompare(b, undefined, { numeric: true }));
+  const out = [];
+  for (const n of entryNames) {
+    const xml = await zip.file(n).async("string");
+    const texts = [...xml.matchAll(/<a:t>([^<]*)<\/a:t>/g)].map((m) => m[1]).join("");
+    out.push(texts);
+  }
+  return out.join("\n\n");
+}
+
+async function officeToText(path) {
+  const ext = extname(path).toLowerCase();
+  const buf = await readFile(path);
+  if (ext === ".docx") return docxToText(buf);
+  if (ext === ".doc") return docToText(buf);
+  if (ext === ".xls") return xlsToText(buf);
+  if (ext === ".xlsx") return xlsxToText(buf);
+  if (ext === ".pptx") return pptxToText(buf);
+  if (ext === ".ppt") return `暂不支持读取旧版 .ppt（请转换为 .pptx）`;
+  if (/\.(txt|md|markdown|log)$/i.test(ext)) return buf.toString("utf8");
+  return `不支持读取该类型：${ext || "未知"}`;
+}
+
+// 声明 apply 需要的 DSH 服务（否则 ctx.tools/ctx.webServer 会报 "without inject"）
+export const inject = ["webServer", "tools"];
+
+async function apply(ctx, config = {}) {
+  // 容错：任何一步失败都只记录、不拖垮 DSH 启动
+  try {
+    await applyInner(ctx, config);
+  } catch (error) {
+    if (ctx && ctx.logger) ctx.logger.error(error instanceof Error ? error : new Error(String(error)));
+    else console.error("[dsh-long-office] apply failed:", error);
+  }
+}
+
+async function applyInner(ctx, config = {}) {
+  const onError = (error) => ctx.logger.error(error instanceof Error ? error : new Error(String(error)));
+  const runScript = (script, args) => new Promise((resolvePromise) => {
+    const child = spawn("python3", [script, ...args], { stdio: ["ignore", "pipe", "pipe"] });
+    let stdout = "", stderr = "";
+    child.stdout.on("data", (d) => { stdout += d.toString(); });
+    child.stderr.on("data", (d) => { stderr += d.toString(); });
+    child.on("error", (error) => resolvePromise({ ok: false, error: String(error), stdout, stderr }));
+    child.on("close", (code) => resolvePromise({ ok: code === 0, code, stdout, stderr }));
+  });
+
+  // 工具：md2docx（从 dsh-long-plugins 搬来）
+  ctx.tools.register(defineTool({
+    name: "md2docx",
+    description: "Convert a Markdown file to a styled Word (.docx) document with a page-number footer (python-docx). Requires python3 + python-docx. Override the script path via config.md2docxScript.",
+    parameters: {
+      input: { type: "string", required: true, description: "Absolute path to the input .md file." },
+      output: { type: "string", description: "Optional absolute path for the output .docx. Defaults to input path with .docx extension." }
+    },
+    output: {
+      schema: { type: "object", additionalProperties: false, properties: { ok: { type: "boolean", required: true }, docxPath: { type: "string" }, error: { type: "string" } } },
+      render: (_args, value) => [{ type: "text", text: value && value.ok === true ? `已生成 Word 文档：${value.docxPath}` : `md2docx 失败：${value?.error ?? "未知错误"}` }]
+    },
+    presentCall: (args) => {
+      const inPath = resolve(String(args.input ?? ""));
+      const outPath = args.output ? resolve(String(args.output)) : inPath.replace(/\.md$/i, ".docx");
+      return { card: "generic", title: "md2docx", kind: "edit", locations: [{ path: outPath }] };
+    },
+    async execute(args) {
+      const inPath = resolve(String(args.input ?? ""));
+      const outPath = args.output ? resolve(String(args.output)) : inPath.replace(/\.md$/i, ".docx");
+      const result = await runScript(config.md2docxScript ?? MD2DOCX_SCRIPT, [inPath, outPath]);
+      if (!result.ok) return { ok: false, error: (result.stderr || result.stdout || String(result.error)).trim() || `md2docx failed (exit ${result.code})` };
+      return { ok: true, docxPath: outPath };
+    }
+  }));
+
+
+  // 工具：md2xlsx（Markdown 表格 → Excel）
+  ctx.tools.register(defineTool({
+    name: "md2xlsx",
+    description: "Convert Markdown tables to an Excel (.xlsx) workbook: one sheet per table (named from the nearest heading), bold header, thin borders, centered, no shading, auto column width, frozen header row; a cell starting with '=' becomes a real formula. Requires python3 + openpyxl. Override the script path via config.md2xlsxScript.",
+    parameters: {
+      input: { type: "string", required: true, description: "Absolute path to the input .md file." },
+      output: { type: "string", description: "Optional absolute path for the output .xlsx. Defaults to input path with .xlsx extension." },
+      overwrite: { type: "boolean", description: "Overwrite an existing output file (default false)." }
+    },
+    output: {
+      schema: { type: "object", additionalProperties: false, properties: { ok: { type: "boolean", required: true }, xlsxPath: { type: "string" }, error: { type: "string" } } },
+      render: (_args, value) => [{ type: "text", text: value && value.ok === true ? `已生成 Excel：${value.xlsxPath}` : `md2xlsx 失败：${value?.error ?? "未知错误"}` }]
+    },
+    presentCall: (args) => {
+      const inPath = resolve(String(args.input ?? ""));
+      const outPath = args.output ? resolve(String(args.output)) : inPath.replace(/\.md$/i, ".xlsx");
+      return { card: "generic", title: "md2xlsx", kind: "edit", locations: [{ path: outPath }] };
+    },
+    async execute(args) {
+      const inPath = resolve(String(args.input ?? ""));
+      const outPath = args.output ? resolve(String(args.output)) : inPath.replace(/\.md$/i, ".xlsx");
+      const scriptArgs = [inPath, outPath];
+      if (args.overwrite === true) scriptArgs.push("--force");
+      const result = await runScript(config.md2xlsxScript ?? MD2XLSX_SCRIPT, scriptArgs);
+      if (!result.ok) return { ok: false, error: (result.stderr || result.stdout || String(result.error)).trim() || `md2xlsx failed (exit ${result.code})` };
+      return { ok: true, xlsxPath: outPath };
+    }
+  }));
+
+  // 工具：md2pptx（Markdown 大纲 → PowerPoint，16:9，带备注/图片/坐标回显）
+  ctx.tools.register(defineTool({
+    name: "md2pptx",
+    description: "Convert a Markdown outline to a 16:9 PowerPoint (.pptx): '# '=cover, '## '=new slide, '- '=bullets, '> 备注：'=speaker notes, '![alt](path \"WxH\")'=image. Requires python3 + python-pptx + Pillow. Override the script path via config.md2pptxScript.",
+    parameters: {
+      input: { type: "string", required: true, description: "Absolute path to the input .md file." },
+      output: { type: "string", description: "Optional absolute path for the output .pptx. Defaults to input path with .pptx extension." },
+      overwrite: { type: "boolean", description: "Overwrite an existing output file (default false)." }
+    },
+    output: {
+      schema: { type: "object", additionalProperties: false, properties: { ok: { type: "boolean", required: true }, pptxPath: { type: "string" }, error: { type: "string" } } },
+      render: (_args, value) => [{ type: "text", text: value && value.ok === true ? `已生成 PPT：${value.pptxPath}` : `md2pptx 失败：${value?.error ?? "未知错误"}` }]
+    },
+    presentCall: (args) => {
+      const inPath = resolve(String(args.input ?? ""));
+      const outPath = args.output ? resolve(String(args.output)) : inPath.replace(/\.md$/i, ".pptx");
+      return { card: "generic", title: "md2pptx", kind: "edit", locations: [{ path: outPath }] };
+    },
+    async execute(args) {
+      const inPath = resolve(String(args.input ?? ""));
+      const outPath = args.output ? resolve(String(args.output)) : inPath.replace(/\.md$/i, ".pptx");
+      const scriptArgs = [inPath, outPath];
+      if (args.overwrite === true) scriptArgs.push("--force");
+      const result = await runScript(config.md2pptxScript ?? MD2PPTX_SCRIPT, scriptArgs);
+      if (!result.ok) return { ok: false, error: (result.stderr || result.stdout || String(result.error)).trim() || `md2pptx failed (exit ${result.code})` };
+      return { ok: true, pptxPath: outPath };
+    }
+  }));
+
+  // 工具：读 Office → 文本/表格（内存、无中间文件、模型按需用）
+  ctx.tools.register(defineTool({
+    name: "office_read",
+    description: "Read an Office/Word/Excel/PowerPoint file (.docx/.doc/.xlsx/.xls/.pptx/.ppt) and return its text/table content in memory (no intermediate files). Use to let the model see/analyze office document content on demand.",
+    parameters: {
+      path: { type: "string", required: true, description: "Absolute path to the office file." }
+    },
+    output: {
+      schema: { type: "object", additionalProperties: false, properties: { ok: { type: "boolean", required: true }, content: { type: "string" }, error: { type: "string" } } },
+      render: (_args, value) => [{ type: "text", text: value && value.ok === true ? (value.content || "(空)") : `office_read 失败：${value?.error ?? "未知错误"}` }]
+    },
+    async execute(args) {
+      try {
+        const path = resolve(String(args.path ?? ""));
+        const content = await officeToText(path);
+        return { ok: true, content: String(content || "").slice(0, 1_000_000) };
+      } catch (error) {
+        return { ok: false, error: String(error && error.message || error) };
+      }
+    }
+  }));
+
+  // 路由：返回 office 文件的内存文本/表格（供前端/工具读取）
+  ctx.effect(() => ctx.webServer.register({
+    kind: "exact",
+    path: "/api/dsh-long-office/content",
+    handler: async (req, res) => {
+      try {
+        if (req.method !== "GET" && req.method !== "HEAD") { res.writeHead(405); res.end(); return; }
+        const rel = (() => { try { return decodeURIComponent(new URL(req.url || "/", "http://dsh.internal").searchParams.get("path") || ""); } catch { return ""; } })();
+        const content = await officeToText(rel);
+        sendJson(res, 200, { ok: true, content });
+      } catch (error) { onError(error); sendJson(res, 500, { ok: false, error: String(error && error.message || error) }); }
+    },
+  }), "dsh-long-office: content route");
+}
+
+export { apply };
+export const name = "dsh-long-office";

package/lib/md2docx.py

@@ -0,0 +1,300 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""Generic Markdown -> styled .docx converter with a page-number footer.
+
+Usage:
+  python3 md2docx.py <input.md> [output.docx]
+
+If output.docx is omitted, it defaults to <input-name>.docx next to the input.
+
+Renders headings (h1-h3), bold/italic inline, tables (pipe syntax),
+ordered/unordered lists, blockquotes, and horizontal rules. Adds a centered
+footer with an auto-updating PAGE field (updates when opened in Word or
+exported to PDF).
+
+Requires: python3 + python-docx (`pip install python-docx`).
+"""
+import re
+import sys
+import os
+
+from docx import Document
+from docx.shared import Pt, RGBColor
+from docx.enum.text import WD_ALIGN_PARAGRAPH
+from docx.enum.table import WD_ALIGN_VERTICAL
+from docx.enum.table import WD_TABLE_ALIGNMENT
+from docx.oxml.ns import qn
+from docx.oxml import OxmlElement
+
+
+def die(msg):
+    print(f"md2docx: {msg}", file=sys.stderr)
+    sys.exit(1)
+
+
+if len(sys.argv) < 2:
+    die("usage: md2docx.py <input.md> [output.docx]")
+
+SRC = os.path.abspath(sys.argv[1])
+if not os.path.isfile(SRC):
+    die(f"input file not found: {SRC}")
+OUT = (
+    os.path.abspath(sys.argv[2])
+    if len(sys.argv) > 2
+    else os.path.splitext(SRC)[0] + ".docx"
+)
+
+with open(SRC, encoding="utf-8") as f:
+    lines = f.read().splitlines()
+
+
+def add_page_number(paragraph):
+    """Insert an auto-updating '第 N 页' page field into a footer paragraph."""
+    run = paragraph.add_run("第 ")
+    set_east_asia(run)
+    fld = OxmlElement("w:fldChar")
+    fld.set(qn("w:fldCharType"), "begin")
+    instr = OxmlElement("w:instrText")
+    instr.set(qn("xml:space"), "preserve")
+    instr.text = " PAGE "
+    sep = OxmlElement("w:fldChar")
+    sep.set(qn("w:fldCharType"), "separate")
+    t = OxmlElement("w:t")
+    t.text = "1"
+    end = OxmlElement("w:fldChar")
+    end.set(qn("w:fldCharType"), "end")
+    r = paragraph.add_run()
+    set_east_asia(r)
+    for el in (fld, instr, sep, t, end):
+        r._r.append(el)
+    run2 = paragraph.add_run(" 页")
+    set_east_asia(run2)
+
+
+doc = Document()
+
+# --- base styles ---
+normal = doc.styles["Normal"]
+normal.font.name = "宋体"
+normal.font.size = Pt(10.5)
+normal._element.rPr.rFonts.set(qn("w:eastAsia"), "宋体")
+# --- 2026-09-14 用户反馈「生成的 docx 行距被压（预览里比我自己放上去的文件紧）」---
+# 根因：python-docx 默认模板的 docDefaults 是 <w:spacing w:line="276" w:lineRule="auto"/>（≈1.15 倍），
+# 而用户自己的文件基本都是 1.5 倍（w:line="360"）。这里统一为 1.5 倍，Word 打开与预览观感一致。
+normal.paragraph_format.line_spacing = 1.5
+
+
+# --- 2026-09-12 用户要求：全篇只用黑色字体、任何底纹都不能有 ---
+normal.font.color.rgb = RGBColor(0, 0, 0)
+for _sn in ("Title", "Heading 1", "Heading 2", "Heading 3", "Heading 4",
+            "Intense Quote", "List Bullet", "List Number", "Table Grid"):
+    try:
+        _st = doc.styles[_sn]
+    except KeyError:
+        continue
+    try:
+        _st.font.color.rgb = RGBColor(0, 0, 0)
+    except Exception:
+        pass
+    try:  # 2026-09-15 用户要求：全部报告/文件用宋体（含标题，避免继承主题 major font）
+        _st.font.name = "宋体"
+        _st.element.rPr.rFonts.set(qn("w:eastAsia"), "宋体")
+    except Exception:
+        pass
+    try:  # 2026-09-14：标题/列表等样式同样统一 1.5 倍行距（否则仍继承 docDefaults 的 1.15 倍）
+        _st.paragraph_format.line_spacing = 1.5
+    except Exception:
+        pass
+    # 去掉该样式自带的底纹（w:shd）
+    for _el in list(_st.element.iter()):
+        if _el.tag == qn("w:shd"):
+            _el.getparent().remove(_el)
+    # 2026-09-15 用户要求：报告不要页眉、标题下不要装饰线 → 去掉样式自带的段落下边框
+    for _el in list(_st.element.iter()):
+        if _el.tag == qn("w:pBdr"):
+            _el.getparent().remove(_el)
+# 清空页眉（只用页脚页码）
+try:
+    for _hp in doc.sections[0].header.paragraphs:
+        for _r in list(_hp.runs):
+            _r._element.getparent().remove(_r._element)
+except Exception:
+    pass
+
+
+def set_east_asia(run):
+    run.font.name = "宋体"
+    run.font.color.rgb = RGBColor(0, 0, 0)   # 2026-09-12：只用黑色
+    r = run._element
+    rPr = r.get_or_add_rPr()
+    rf = rPr.find(qn("w:rFonts"))
+    if rf is None:
+        rf = OxmlElement("w:rFonts")
+        rPr.append(rf)
+    rf.set(qn("w:ascii"), "Times New Roman")     # 西文/数字：Times New Roman（宋体空格是全角）
+    rf.set(qn("w:hAnsi"), "Times New Roman")
+    rf.set(qn("w:eastAsia"), "宋体")
+
+
+def add_runs_with_bold(par, text):
+    """2026-09-15 用户要求：**正文一律不加粗，只有标题加粗**。
+
+    md 里的 `**…**` / `*…*` 仅作标记清理，不再转成加粗/斜体
+    （标题由 Heading 样式自带加粗；表格表头由 flush_table 单独置粗）。
+    """
+    r = par.add_run(re.sub(r"\*+", "", text))
+    set_east_asia(r)
+
+
+def add_body_paragraph(text, style=None):
+    p = doc.add_paragraph(style=style)
+    add_runs_with_bold(p, text)
+    return p
+
+
+def flush_table():
+    global table_rows, in_table
+    if not table_rows:
+        in_table = False
+        return
+    ncols = max(len(r) for r in table_rows)
+    tbl = doc.add_table(rows=len(table_rows), cols=ncols)
+    tbl.style = "Table Grid"   # 2026-09-12：原来 Light Grid Accent 1 带底纹/彩色框线
+    tbl.alignment = WD_TABLE_ALIGNMENT.CENTER
+    for ri, row in enumerate(table_rows):
+        for ci in range(ncols):
+            cell = tbl.cell(ri, ci)
+            cell.text = ""
+            cp = cell.paragraphs[0]
+            text = row[ci] if ci < len(row) else ""
+            add_runs_with_bold(cp, text)
+            if ri == 0:
+                for run in cp.runs:
+                    run.bold = True
+    tbl.autofit = True
+    table_rows = []
+    in_table = False
+
+
+table_rows = []
+in_table = False
+
+i = 0
+while i < len(lines):
+    stripped = lines[i].strip()
+    if not stripped:
+        i += 1
+        continue
+    if re.fullmatch(r"-{3,}", stripped):
+        flush_table()
+        doc.add_paragraph()
+        i += 1
+        continue
+    if in_table and "|" in stripped and re.fullmatch(r"\|?[\s:|-]+\|?", stripped):
+        i += 1
+        continue
+    if stripped.startswith("|"):
+        in_table = True
+        table_rows.append([c.strip() for c in stripped.strip("|").split("|")])
+        i += 1
+        continue
+    if in_table:
+        flush_table()
+    if stripped.startswith("### "):
+        h = doc.add_heading(level=3)
+        add_runs_with_bold(h, stripped[4:])
+        i += 1
+        continue
+    if stripped.startswith("## "):
+        h = doc.add_heading(level=2)
+        add_runs_with_bold(h, stripped[3:])
+        i += 1
+        continue
+    if stripped.startswith("# "):
+        h = doc.add_heading(level=1)
+        h.alignment = WD_ALIGN_PARAGRAPH.CENTER          # ★文档标题居中（用户要求）
+        add_runs_with_bold(h, stripped[2:])
+        i += 1
+        continue
+    if stripped.startswith("> "):
+        p = doc.add_paragraph(style="Intense Quote")
+        add_runs_with_bold(p, stripped[2:] + " ")
+        i += 1
+        continue
+    m = re.match(r"^(\d+)\.\s+(.*)$", stripped)
+    if m:
+        p = doc.add_paragraph(style="List Number")
+        add_runs_with_bold(p, m.group(2))
+        i += 1
+        continue
+    if stripped.startswith("- "):
+        p = doc.add_paragraph(style="List Bullet")
+        add_runs_with_bold(p, stripped[2:])
+        i += 1
+        continue
+    p = doc.add_paragraph()
+    add_runs_with_bold(p, stripped)
+    i += 1
+
+if in_table:
+    flush_table()
+
+# --- footer with page number field ---
+footer = doc.sections[0].footer
+fp = footer.paragraphs[0]
+fp.alignment = WD_ALIGN_PARAGRAPH.CENTER
+add_page_number(fp)
+# 2026-09-15 用户要求：页码字体与正文一致（宋体）、居中
+for _r in fp.runs:
+    _r.font.name = "宋体"
+    _r.font.color.rgb = RGBColor(0, 0, 0)
+    _rp = _r._element.get_or_add_rPr()
+    _rp.get_or_add_rFonts().set(qn("w:eastAsia"), "宋体")
+try:
+    _fs = fp.style
+    _fs.font.name = "宋体"
+    _fs.element.rPr.rFonts.set(qn("w:eastAsia"), "宋体")
+except Exception:
+    pass
+
+# --- 2026-09-15 用户要求：**所有生成文件**都必须符合 skill `doc-heading-format` ---
+# 在源头统一：正文首行缩进 2 字符、表格内容居中（水平＋垂直）、全黑字、无底纹、行距 1.5。
+# 放在这里，任何技能/任何调用 md2docx 得到的 docx 都自动合规，不依赖调用方是否记得。
+def apply_doc_format(doc):
+    for p in doc.paragraphs:
+        st = (p.style.name or "")
+        is_head = st.startswith(("Heading", "Title")) or p.alignment is not None
+        pPr = p._p.get_or_add_pPr()
+        for _shd in pPr.findall(qn("w:shd")):
+            pPr.remove(_shd)
+        for r in p.runs:
+            r.font.color.rgb = RGBColor(0, 0, 0)
+        p.paragraph_format.line_spacing = 1.5
+        p.paragraph_format.space_after = Pt(6)
+        # 只有普通正文段落才首行缩进（标题/列表项/引用不缩进）
+        if (not is_head) and p.text.strip() and st in ("Normal", ""):
+            ind = pPr.get_or_add_ind()
+            ind.set(qn("w:firstLineChars"), "200")   # 2 字符（中文习惯）
+            ind.set(qn("w:firstLine"), "420")
+    for t in doc.tables:
+        for row in t.rows:
+            for c in row.cells:
+                try:
+                    c.vertical_alignment = WD_ALIGN_VERTICAL.CENTER
+                except Exception:
+                    pass
+                tcPr = c._tc.get_or_add_tcPr()
+                for _shd in tcPr.findall(qn("w:shd")):
+                    tcPr.remove(_shd)
+                for p in c.paragraphs:
+                    p.alignment = WD_ALIGN_PARAGRAPH.CENTER   # ★表格内容一律居中
+                    p.paragraph_format.line_spacing = 1.15
+                    p.paragraph_format.space_after = Pt(2)
+                    for r in p.runs:
+                        r.font.color.rgb = RGBColor(0, 0, 0)
+
+
+apply_doc_format(doc)
+
+doc.save(OUT)
+print("Saved:", OUT)

package/lib/md2pptx.py

@@ -0,0 +1,276 @@
+#!/usr/bin/env python3
+"""md2pptx.py — Markdown 大纲 → PowerPoint(.pptx)，16:9
+
+完全自研（只依赖 python-pptx + Pillow），排版口径沿用本机硬规则：
+  · 白底、**无底纹/无花哨填充**、文字全黑；中文宋体、纯西文 Times New Roman
+  · 16:9（13.333 × 7.5 英寸）；标题 28pt 粗、正文 18pt、备注 12pt
+  · **演讲者备注**：`> 备注：...` 行写进当前页 notes
+  · **图片**：`![说明](路径)` 默认按原图比例放入内容区；`![说明](路径 "4x3")` 指定英寸宽×高
+  · **坐标回显**：生成后打印每页每个元素的落点（英寸坐标 + 画布尺寸 + 文本线框）
+
+Markdown 约定（自上而下）:
+  # 标题              → 封面（副标题取 `> ...`）
+  ## 小节标题         → 新的一页（内容页）
+  - 要点 /   - 子要点  → 项目符号（1/2 级）
+  > 备注：xxx         → 当前页备注（封面页则作副标题）
+  ![说明](图片路径)    → 当前页插图
+  ---                → 换页（无标题）
+  （普通段落）         → 当前页正文段落
+
+用法:
+  python3 md2pptx.py <input.md> [output.pptx] [--force] [--echo-json out.json]
+  默认输出同名 .pptx；默认不覆盖已存在文件，--force 才覆盖。
+"""
+import json
+import os
+import re
+import sys
+
+from PIL import Image
+from pptx import Presentation
+from pptx.dml.color import RGBColor
+from pptx.enum.text import PP_ALIGN
+from pptx.oxml.ns import qn
+from pptx.util import Inches, Pt
+
+EAST, LATIN = "宋体", "Times New Roman"
+BLACK = RGBColor(0, 0, 0)
+SLIDE_W, SLIDE_H = 13.333, 7.5
+MARGIN, TITLE_H = 0.6, 1.2
+TITLE_PT, BODY_PT, NOTE_PT = 28, 18, 12
+BULLET_CHARS = "•", "–"
+ASCII_RE = re.compile(r"^[\x00-\x7F]*$")
+SEG_RE = re.compile(r"[\x00-\x7F]+|[^\x00-\x7F]+")     # 按 ASCII / 非 ASCII 分段，西文走 TNR
+IMG_RE = re.compile(r"^!\[(?P<alt>[^\]]*)\]\((?P<path>[^)\s]+)(?:\s+\"(?P<size>[0-9.]+x[0-9.]+)\")?\)\s*$")
+NOTE_RE = re.compile(r"^>\s*(?:备注|Notes?|note)\s*[:：]?\s*(.*)$", re.I)
+QUOTE_RE = re.compile(r"^>\s*(.*)$")
+
+
+def parse(md_text):
+    """→ [ {title, subtitle, bullets:[(level,text)], images:[(path,alt,w,h)], paragraphs:[str], notes:str} ]"""
+    slides, cur = [], None
+
+    def new(title=None):
+        s = {"title": title, "subtitle": None, "bullets": [], "images": [],
+             "paragraphs": [], "notes": ""}
+        slides.append(s)
+        return s
+
+    for raw in md_text.splitlines():
+        line = raw.rstrip()
+        s = line.strip()
+        if not s:
+            continue
+        m_img = IMG_RE.match(s)
+        if s == "---":
+            new()
+            continue
+        if s.startswith("# "):
+            cur = new(s[2:].strip())          # 封面
+            continue
+        if s.startswith("## "):
+            cur = new(s[3:].strip())          # 内容页
+            continue
+        if cur is None:
+            cur = new()
+        m_note = NOTE_RE.match(s)
+        if m_note:
+            if cur["title"] and not cur["bullets"] and not cur["paragraphs"] and not cur["notes"]:
+                cur["subtitle"] = m_note.group(1).strip()   # 封面页的 > 备注： → 副标题
+            else:
+                cur["notes"] = (cur["notes"] + "\n" + m_note.group(1).strip()).strip()
+            continue
+        m_quote = QUOTE_RE.match(s)
+        if m_quote:
+            cur["notes"] = (cur["notes"] + "\n" + m_quote.group(1).strip()).strip()
+            continue
+        if m_img:
+            size = m_img.group("size")
+            w, h = (float(x) for x in size.split("x")) if size else (None, None)
+            cur["images"].append((m_img.group("path"), m_img.group("alt"), w, h))
+            continue
+        m_b = re.match(r"^(\s*)[-*+]\s+(.*)$", raw)
+        if m_b:
+            level = 1 if len(m_b.group(1)) < 2 else 2
+            cur["bullets"].append((level, m_b.group(2).strip()))
+            continue
+        cur["paragraphs"].append(s)
+    return [s for s in slides if any([s["title"], s["bullets"], s["images"], s["paragraphs"]])]
+
+
+def style_run(run, size, bold=False):
+    run.font.size = Pt(size)
+    run.font.bold = bold
+    run.font.color.rgb = BLACK
+    txt = run.text or ""
+    run.font.name = LATIN if ASCII_RE.fullmatch(txt) and txt else EAST
+    # 东亚字体显式设置（python-pptx 的 font.name 只管 latin）
+    rPr = run._r.get_or_add_rPr()
+    for tag in ("a:ea", "a:cs"):
+        el = rPr.find(qn(tag))
+        if el is None:
+            el = rPr.makeelement(qn(tag), {})
+            rPr.append(el)
+        el.set("typeface", EAST)
+
+
+def add_styled_runs(paragraph, text, size, bold=False):
+    """按'纯西文段／中日韩段'拆成多个 run：西文 Times New Roman、中文宋体（用户硬规则）"""
+    wrote = False
+    for seg in SEG_RE.findall(text):
+        if not seg:
+            continue
+        run = paragraph.add_run()
+        run.text = seg
+        style_run(run, size, bold)
+        wrote = True
+    if not wrote:
+        run = paragraph.add_run()
+        run.text = ""
+        style_run(run, size, bold)
+
+
+def add_textbox(slide, left, top, width, height):
+    tb = slide.shapes.add_textbox(Inches(left), Inches(top), Inches(width), Inches(height))
+    tf = tb.text_frame
+    tf.word_wrap = True
+    return tf
+
+
+def fill_title(slide, title, subtitle):
+    tf = add_textbox(slide, MARGIN, 0.35, SLIDE_W - 2 * MARGIN, TITLE_H)
+    add_styled_runs(tf.paragraphs[0], title or "", TITLE_PT, bold=True)
+    if subtitle:
+        add_styled_runs(tf.add_paragraph(), subtitle, 14)
+
+
+def picture_size(path, w, h):
+    with Image.open(path) as im:
+        iw, ih = im.size
+    ratio = iw / ih if ih else 1.0
+    if w and h:
+        return w, h
+    box_w, box_h = 5.4, 4.6
+    if not w and not h:
+        w = box_w
+        h = w / ratio
+        if h > box_h:
+            h = box_h
+            w = h * ratio
+    elif w and not h:
+        h = w / ratio
+    else:
+        w = h * ratio
+    return w, h
+
+
+def build(md_text, echo_json=None):
+    prs = Presentation()
+    prs.slide_width, prs.slide_height = Inches(SLIDE_W), Inches(SLIDE_H)
+    blank = prs.slide_layouts[6]
+    geometry = []
+
+    for spec in parse(md_text):
+        slide = prs.slides.add_slide(blank)
+        si = len(prs.slides._sldIdLst)          # 1-based 页码
+        is_cover = spec["subtitle"] is not None or (spec["title"] and not spec["bullets"]
+                                                   and not spec["images"] and not spec["paragraphs"]
+                                                   and not spec["notes"])
+        if spec["title"]:
+            if is_cover:
+                tf = add_textbox(slide, MARGIN, 2.6, SLIDE_W - 2 * MARGIN, 1.6)
+                p = tf.paragraphs[0]
+                p.alignment = PP_ALIGN.CENTER
+                add_styled_runs(p, spec["title"], 40, bold=True)
+                if spec["subtitle"]:
+                    p2 = tf.add_paragraph()
+                    p2.alignment = PP_ALIGN.CENTER
+                    add_styled_runs(p2, spec["subtitle"], 16)
+            else:
+                fill_title(slide, spec["title"], None)
+
+        has_img = bool(spec["images"])
+        text_w = 6.2 if has_img else SLIDE_W - 2 * MARGIN
+        top = 1.75 if spec["title"] and not is_cover else 1.2
+
+        if spec["bullets"] or spec["paragraphs"]:
+            tf = add_textbox(slide, MARGIN, top, text_w, SLIDE_H - top - MARGIN)
+            first = True
+            for level, text in spec["bullets"]:
+                p = tf.paragraphs[0] if first else tf.add_paragraph()
+                first = False
+                p.level = min(level - 1, 4)
+                p.space_after = Pt(6)
+                bullet = BULLET_CHARS[min(level, 2) - 1]
+                add_styled_runs(p, f"{bullet} {text}", BODY_PT)
+            for text in spec["paragraphs"]:
+                p = tf.paragraphs[0] if first else tf.add_paragraph()
+                first = False
+                add_styled_runs(p, text, BODY_PT)
+
+        for idx, (path, alt, w, h) in enumerate(spec["images"]):
+            if not os.path.isfile(path):
+                print(f"  警告：图片不存在，已跳过 → {path}", file=sys.stderr)
+                continue
+            w, h = picture_size(path, w, h)
+            left = SLIDE_W - MARGIN - w if has_img else MARGIN
+            pic_top = max(top, (SLIDE_H - h) / 2)
+            pic = slide.shapes.add_picture(path, Inches(left), Inches(pic_top), Inches(w), Inches(h))
+            pic.name = alt or os.path.basename(path)
+
+        if spec["notes"]:
+            tf = slide.notes_slide.notes_text_frame
+            tf.text = spec["notes"]
+            for p in tf.paragraphs:
+                for r in p.runs:
+                    style_run(r, NOTE_PT)
+
+        for sh in slide.shapes:
+            kind = str(sh.shape_type).split()[0] if sh.shape_type is not None else "TEXT_BOX"
+            text = sh.text_frame.text.replace("\n", " / ")[:60] if sh.has_text_frame else (
+                sh.name if kind == "PICTURE" else "")
+            geometry.append({
+                "slide": si,
+                "shape": kind,
+                "name": sh.name,
+                "left_in": round(sh.left / 914400, 3), "top_in": round(sh.top / 914400, 3),
+                "width_in": round(sh.width / 914400, 3), "height_in": round(sh.height / 914400, 3),
+                "text": text,
+            })
+    if echo_json:
+        with open(echo_json, "w", encoding="utf-8") as fh:
+            json.dump({"canvas_in": [SLIDE_W, SLIDE_H], "slides": geometry}, fh, ensure_ascii=False, indent=2)
+    return prs, geometry
+
+
+def main():
+    argv = [a for a in sys.argv[1:] if not a.startswith("--")]
+    force = "--force" in sys.argv
+    echo = None
+    if "--echo-json" in sys.argv:
+        echo = sys.argv[sys.argv.index("--echo-json") + 1]
+    if not argv:
+        print("用法: md2pptx.py <input.md> [output.pptx] [--force] [--echo-json out.json]", file=sys.stderr)
+        return 2
+    src = os.path.abspath(argv[0])
+    out = os.path.abspath(argv[1]) if len(argv) > 1 else re.sub(r"\.md$", "", src, flags=re.I) + ".pptx"
+    if os.path.exists(out) and not force:
+        print(f"已存在，未覆盖（加 --force 覆盖）: {out}", file=sys.stderr)
+        return 3
+    if not os.path.isfile(src):
+        print(f"找不到输入: {src}", file=sys.stderr)
+        return 2
+    with open(src, encoding="utf-8") as fh:
+        prs, geo = build(fh.read(), echo)
+    prs.save(out)
+    print("Saved:", out)
+    print(f"  画布 {SLIDE_W}x{SLIDE_H} 英寸（16:9），共 {len(prs.slides._sldIdLst)} 页")
+    cur = None
+    for g in geo:
+        print("  [%s] %-12s left=%.2f top=%.2f w=%.2f h=%.2f in  %s" % (
+            g["slide"], g["shape"][:12], g["left_in"], g["top_in"], g["width_in"], g["height_in"], g["text"]))
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

package/lib/md2xlsx.py

@@ -0,0 +1,171 @@
+#!/usr/bin/env python3
+"""md2xlsx.py — Markdown 表格 → Excel(.xlsx)
+
+排版口径与 dsh-office-reader 的 md2docx 一致（用户硬规则）：
+  · 中文宋体 / 纯西文单元格 Times New Roman；全黑字；**无底纹（不加填充色）**
+  · 表格：细边框、内容水平+垂直居中、表头加粗、列宽按内容自适应、首行冻结
+  · 以 `=` 开头的单元格写成**真正的 Excel 公式**；数字串转数值
+  · 每个 Markdown 表格 → 一个 sheet，表名取该表格上方最近的标题（无则 Sheet1/2…）
+
+用法:
+  python3 md2xlsx.py <input.md> [output.xlsx] [--force]
+  默认输出同名 .xlsx；默认**不覆盖**已存在文件（防误覆盖），--force 才覆盖。
+"""
+import os
+import re
+import sys
+
+from openpyxl import Workbook
+from openpyxl.styles import Alignment, Border, Font, PatternFill, Side
+from openpyxl.utils import get_column_letter
+
+EAST = "宋体"
+LATIN = "Times New Roman"
+THIN = Side(style="thin", color="000000")
+BORDER = Border(left=THIN, right=THIN, top=THIN, bottom=THIN)
+CENTER = Alignment(horizontal="center", vertical="center", wrap_text=True)
+HEAD_FONT_EA = Font(name=EAST, bold=True, color="FF000000")
+BODY_FONT_EA = Font(name=EAST, color="FF000000")
+HEAD_FONT_LA = Font(name=LATIN, bold=True, color="FF000000")
+BODY_FONT_LA = Font(name=LATIN, color="FF000000")
+
+NUM_RE = re.compile(r"^-?\d+(?:\.\d+)?%?$")
+ASCII_RE = re.compile(r"^[\x00-\x7F]*$")
+BAD_SHEET = re.compile(r"[\\/*?:\[\]]")
+
+
+def is_sep_row(cells):
+    """| --- | :--: | 这种分隔行"""
+    return all(re.fullmatch(r":?-{2,}:?", c.strip()) for c in cells if c.strip() != "") and any(
+        c.strip() for c in cells
+    )
+
+
+def split_row(line):
+    raw = line.strip()
+    if raw.startswith("|"):
+        raw = raw[1:]
+    if raw.endswith("|"):
+        raw = raw[:-1]
+    return [c.strip() for c in raw.split("|")]
+
+
+def parse_tables(md_text):
+    """→ [(sheet_name, rows)]  rows = [[cell, ...], ...]"""
+    out, cur, heading, pending_heading = [], [], None, None
+    in_table = False
+    for line in md_text.splitlines():
+        s = line.strip()
+        if s.startswith("#"):
+            pending_heading = s.lstrip("#").strip() or None
+            if in_table:
+                in_table = False
+            continue
+        if s.startswith("|") and s.count("|") >= 2:
+            cells = split_row(s)
+            if not in_table:
+                in_table = True
+                cur = []
+                heading = pending_heading
+            if is_sep_row(cells):
+                continue
+            cur.append(cells)
+        else:
+            if in_table and cur:
+                out.append((heading, cur))
+                cur, in_table = [], False
+    if in_table and cur:
+        out.append((heading, cur))
+    return out
+
+
+def clean_sheet_name(name, used, idx):
+    base = BAD_SHEET.sub("-", (name or "").strip()) or f"Sheet{idx}"
+    base = base[:31]
+    cand, n = base, 2
+    while cand in used:
+        suffix = f"({n})"
+        cand = base[: 31 - len(suffix)] + suffix
+        n += 1
+    used.add(cand)
+    return cand
+
+
+def put_cell(ws, r, c, raw, bold):
+    val = raw
+    if val.startswith("="):
+        ws.cell(row=r, column=c, value=val)          # 真公式
+    elif NUM_RE.fullmatch(val):
+        try:
+            v = float(val[:-1]) / 100 if val.endswith("%") else float(val)
+            ws.cell(row=r, column=c, value=v)
+        except ValueError:
+            ws.cell(row=r, column=c, value=val)
+    else:
+        ws.cell(row=r, column=c, value=val)
+    cell = ws.cell(row=r, column=c)
+    ascii_only = bool(ASCII_RE.fullmatch(val)) and val != ""
+    if bold:
+        cell.font = HEAD_FONT_LA if ascii_only else HEAD_FONT_EA
+    else:
+        cell.font = BODY_FONT_LA if ascii_only else BODY_FONT_EA
+    cell.alignment = CENTER
+    cell.border = BORDER
+    # 清掉任何填充（用户硬规则：不许有底纹）
+    cell.fill = PatternFill(fill_type=None)
+
+
+def write_sheet(wb, name, rows, idx):
+    ws = wb.create_sheet(title=name)
+    for ri, row in enumerate(rows, start=1):
+        for ci, raw in enumerate(row, start=1):
+            put_cell(ws, ri, ci, raw, bold=(ri == 1))
+    # 列宽自适应（中文按 2 个字符宽估）
+    for ci in range(1, max(len(r) for r in rows) + 1):
+        widest = 0
+        for row in rows:
+            if ci <= len(row):
+                txt = row[ci - 1]
+                w = sum(2 if ord(ch) > 127 else 1 for ch in txt)
+                widest = max(widest, w)
+        ws.column_dimensions[get_column_letter(ci)].width = min(max(widest + 2, 6), 60)
+    # 行高与冻结表头
+    for ri in range(1, len(rows) + 1):
+        ws.row_dimensions[ri].height = 20
+    ws.freeze_panes = "A2"
+    return ws
+
+
+def main():
+    args = [a for a in sys.argv[1:] if not a.startswith("--")]
+    force = "--force" in sys.argv
+    if not args:
+        print("用法: md2xlsx.py <input.md> [output.xlsx] [--force]", file=sys.stderr)
+        return 2
+    src = os.path.abspath(args[0])
+    out = os.path.abspath(args[1]) if len(args) > 1 else re.sub(r"\.md$", "", src, flags=re.I) + ".xlsx"
+    if os.path.exists(out) and not force:
+        print(f"已存在，未覆盖（加 --force 覆盖）: {out}", file=sys.stderr)
+        return 3
+    if not os.path.isfile(src):
+        print(f"找不到输入: {src}", file=sys.stderr)
+        return 2
+
+    with open(src, encoding="utf-8") as fh:
+        tables = parse_tables(fh.read())
+    if not tables:
+        print("输入里没有 Markdown 表格（| … | 且含分隔行）", file=sys.stderr)
+        return 4
+
+    wb = Workbook()
+    wb.remove(wb.active)
+    used = set()
+    for i, (heading, rows) in enumerate(tables, start=1):
+        write_sheet(wb, clean_sheet_name(heading, used, i), rows, i)
+    wb.save(out)
+    print("Saved:", out)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

package/package.json

@@ -0,0 +1,56 @@
+{
+  "name": "dsh-long-office",
+  "version": "0.1.2",
+  "description": "DSH Office read + generate: read Word/Excel/PowerPoint (docx/xlsx/pptx and legacy doc/xls) into text/tables in memory (no intermediate files, model on-demand); generate Word/Excel/PowerPoint from Markdown via md2docx / md2xlsx / md2pptx. | DSH 插件：Office 读取 + 生成 —— 内存读取 Word/Excel/PPT（含旧版 doc/xls）为文本/表格；Markdown 生成 Word/Excel/PPT（md2docx / md2xlsx / md2pptx）。",
+  "type": "module",
+  "license": "MIT",
+  "main": "lib/index.js",
+  "exports": {
+    ".": "./lib/index.js",
+    "./client": "./client/client.js",
+    "./package.json": "./package.json"
+  },
+  "files": [
+    "lib/index.js",
+    "lib/md2docx.py",
+    "lib/md2xlsx.py",
+    "lib/md2pptx.py",
+    "client",
+    "cordis.patch.yml",
+    "dsh.plugin.json"
+  ],
+  "dependencies": {
+    "mammoth": "^1.12.1",
+    "exceljs": "^4.4.0",
+    "xlsx": "^0.18.5",
+    "word-extractor": "^1.0.4",
+    "jszip": "^3.10.1"
+  },
+  "peerDependencies": {
+    "@deepseek-ai/dsh-tools": "*"
+  },
+  "peerDependenciesMeta": {
+    "@deepseek-ai/dsh-tools": {
+      "optional": true
+    }
+  },
+  "dsh": {
+    "bundle": {
+      "patch": "./cordis.patch.yml"
+    },
+    "client": {
+      "inject": [
+        "@deepseek-ai/dsh-client-connection",
+        "@deepseek-ai/dsh-client-runtime",
+        "@deepseek-ai/dsh-client-locale",
+        "@deepseek-ai/dsh-client-ui-slots",
+        "@deepseek-ai/dsh-client-ui-settings",
+        "@deepseek-ai/dsh-client-ui-theme",
+        "@deepseek-ai/dsh-client-ui-conversation",
+        "@deepseek-ai/dsh-client-ui-input-trigger",
+        "@deepseek-ai/dsh-client-ui-layout"
+      ],
+      "platform": "web"
+    }
+  }
+}