@paircode/tool-office 1.0.1 → 1.0.3

package/index.js

@@ -1,8 +1,10 @@
 // ═══════════════════════════════════════════════════════════════
 // tool-office — 办公文档（csv_read/csv_write/json_to_table/table_stats/text_report/word_read）
 //
-// 生成来源（2026-08-16）：内置 Go 工具组 → 磁盘外置插件（tool_plugin_gen.go
-// 自动生成，schema 完整外置拷贝）。api 声明在插件，execute 调 ctx.binary 复用本插件目录 bin/ 下的独立二进制（源码 plugins-src/plugins/<name>/，改实现重编译即更换）。
+// ★ 2026-08-22 JS 原生化（3/11）：csv_read/csv_write/json_to_table 实现
+//   全在插件内（ctx.fs + JS 解析/渲染）；其余（table_stats/text_report/
+//   word_read/word_write/read_xlsx/write_xlsx/read_pdf/markdown_to_html）
+//   保留依赖 bin/tool-office.exe（ctx.binary.exec 桥接）。
 // 工具清单：csv_read、csv_write、json_to_table、table_stats、text_report、word_read、word_write、read_xlsx、write_xlsx、read_pdf、markdown_to_html
 // ═══════════════════════════════════════════════════════════════
 const tools = [
@@ -262,8 +264,8 @@ const tools = [
   },
   {
     "name": "read_pdf",
-    "description": "提取 PDF 文件的文本内容。先尝试纯文本提取（解析 PDF 流对象）；如提取结果为空或内容极少（\u003c50 字符），自动调用 pdftoppm/mutool 将页面渲染为图片 + Tesseract OCR 识别文字。page 指定页码（从 1 开始，默认全部）；limit 限制返回字符数（默认 10000）。",
-    "usageGuide": "提取 PDF 文本内容。先尝试纯文本提取，失败则回退 OCR。比手动打开 PDF 复制更高效（直接提取到上下文）。",
+    "description": "提取 PDF 文件的文本内容（解析 PDF 流对象，支持纯文本 PDF）。扫描/图片型 PDF（无嵌入文本）无法提取，返回说明。page 指定页码（从 1 开始，默认全部）；limit 限制返回字符数（默认 10000）。",
+    "usageGuide": "提取 PDF 文件文本内容（解析 PDF 流对象，纯文本 PDF）。page 指定页码；limit 限制返回字符数。扫描/图片型 PDF 无嵌入文本时返回提示。",
     "parameters": {
       "properties": {
         "limit": {
@@ -314,9 +316,564 @@ const tools = [
   }
 ];
 
+// ─── JS 原生化实现（csv_read/csv_write/json_to_table） ─────────
+
+// CSV 解析（对齐 Go encoding/csv LazyQuotes 语义：字段开头引号才进入引号模式）。
+function readCSV(text, delim) {
+  const rows = []
+  let row = [], field = '', inQuotes = false, fieldStart = true
+  for (let i = 0; i < text.length; i++) {
+    const c = text[i]
+    if (inQuotes) {
+      if (c === '"') {
+        if (text[i + 1] === '"') { field += '"'; i++ }
+        else inQuotes = false
+      } else field += c
+    } else if (fieldStart && c === '"') {
+      inQuotes = true
+    } else if (c === delim) {
+      row.push(field); field = ''; fieldStart = true
+    } else if (c === '\n') {
+      row.push(field); rows.push(row); row = []; field = ''; fieldStart = true
+    } else if (c === '\r') {
+      // 忽略（\r\n 随 \n 处理；单独 \r 作普通字符保留）
+    } else {
+      field += c; fieldStart = false
+    }
+  }
+  if (field !== '' || row.length > 0) { row.push(field); rows.push(row) }
+  return rows
+}
+
+// 生成 CSV（转义：含引号/逗号/换行的字段加引号包裹）。
+function writeCSV(records, delim) {
+  return records.map(row =>
+    row.map(cell => {
+      const s = String(cell == null ? '' : cell)
+      return /["\n\r]/.test(s) || s.includes(delim) ? '"' + s.replace(/"/g, '""') + '"' : s
+    }).join(delim)
+  ).join('\n') + '\n'
+}
+
+function csvDelim(d) {
+  const s = String(d || '').trim().toLowerCase()
+  return (s === 'tab' || s === '\t' || s === '制表符') ? '\t' : ','
+}
+
+function parseColIndex(spec, maxCols) {
+  const s = String(spec || '').trim()
+  if (!s) return null
+  const idx = []
+  for (const part of s.split(',')) {
+    const p = part.trim()
+    const i = Number(p)
+    if (Number.isInteger(i) && i >= 0 && i < maxCols) idx.push(i)
+  }
+  return idx
+}
+
+function padRight(s, n) {
+  s = String(s || '')
+  return s.length >= n ? s : s + ' '.repeat(n - s.length)
+}
+
+// writeMarkdownTable 渲染二维数组为 Markdown 表格（colIdx 为 null 时全列）。
+function writeMarkdownTable(records, colIdx) {
+  if (!records || records.length === 0) return ''
+  const selectCols = (row) => {
+    if (colIdx == null) return row
+    return colIdx.map(idx => (idx < row.length ? row[idx] : ''))
+  }
+  const header = selectCols(records[0])
+  const widths = header.map(h => String(h || '').length)
+  for (const row of records.slice(1)) {
+    const cols = selectCols(row)
+    cols.forEach((c, i) => {
+      if (i < widths.length && String(c || '').length > widths[i]) widths[i] = String(c || '').length
+    })
+    while (widths.length < cols.length) widths.push(0)
+  }
+  while (header.length < widths.length) header.push('')
+  let out = '| '
+  header.forEach((h, i) => { out += padRight(h, widths[i]) + ' | ' })
+  out += '\n| '
+  widths.forEach(w => { out += '-'.repeat(Math.max(w, 3)) + ' | ' })
+  out += '\n'
+  for (const row of records.slice(1)) {
+    const cols = selectCols(row)
+    out += '| '
+    cols.forEach((c, i) => { out += (i < widths.length ? padRight(c, widths[i]) : String(c || '')) + ' | ' })
+    out += '\n'
+  }
+  return out
+}
+
+// parseMarkdownTable 从 Markdown 文本提取表格数据（简单实现，同 Go 版）。
+function parseMarkdownTable(text) {
+  const records = []
+  let inTable = false
+  for (const ln of String(text).split('\n')) {
+    const line = ln.trim()
+    if (!line || !line.startsWith('|')) {
+      if (inTable) break
+      continue
+    }
+    const stripped = line.replace(/-/g, '').replace(/\|/g, '').replace(/ /g, '')
+    if (stripped === '') continue
+    inTable = true
+    let row = line.split('|').map(p => p.trim())
+    if (row.length > 0 && row[0] === '') row = row.slice(1)
+    if (row.length > 0 && row[row.length - 1] === '') row = row.slice(0, -1)
+    if (row.length > 0) records.push(row)
+  }
+  return records
+}
+
+// csv_read：读 CSV/TSV → Markdown 表格
+function csvRead(ctx, args) {
+  const path = String(args.path || '').trim()
+  if (!path) throw new Error('path 不能为空')
+  const data = ctx.fs.readFile(path)
+  if (data.length > 100 << 20) throw new Error('文件超过 100MB，请缩小范围后用 search_content 搜索特定内容')
+  const delim = csvDelim(args.delimiter)
+  let limit = Number(args.limit || 100)
+  let offset = Number(args.offset || 0)
+  const colSpec = String(args.columns || '').trim()
+  const records = readCSV(data, delim)
+  if (records.length === 0) return '（空文件）'
+  const colIdx = parseColIndex(colSpec, records[0].length)
+  if (colSpec && (!colIdx || colIdx.length === 0)) throw new Error('无效的 columns 参数: "' + colSpec + '"，应为逗号分隔的列索引（从 0 开始）')
+  const total = records.length
+  if (offset > 0 && offset < records.length) records.splice(0, offset)
+  else if (offset >= records.length) return '（offset 超出文件行数）'
+  if (limit > 0 && limit < records.length) records.length = limit
+  let out = '**CSV 文件**: `' + path + '` · 共 ' + total + ' 行 × ' + records[0].length + ' 列'
+  if (limit > 0 && total > limit) out += ' · 显示前 ' + limit + ' 行'
+  if (offset > 0) out += ' · 跳过 ' + offset + ' 行'
+  if (colSpec) out += ' · 显示列 [' + colSpec + ']'
+  out += '\n\n' + writeMarkdownTable(records, colIdx)
+  return out
+}
+
+// csv_write：表格数据 → CSV/TSV 文件
+function csvWrite(ctx, args) {
+  const path = String(args.path || '').trim()
+  if (!path) throw new Error('path 不能为空')
+  const dataStr = String(args.data || '')
+  const delim = csvDelim(args.delimiter)
+  let records = null
+  try { records = JSON.parse(dataStr) } catch { records = parseMarkdownTable(dataStr) }
+  if (!Array.isArray(records) || records.length === 0 || !Array.isArray(records[0])) throw new Error('data 格式无效：无法解析为 JSON 二维数组或 Markdown 表格')
+  const headerJSON = String(args.header || '').trim()
+  if (headerJSON) {
+    let header
+    try { header = JSON.parse(headerJSON) } catch { throw new Error('header JSON 解析失败') }
+    if (!Array.isArray(header)) throw new Error('header 必须是 JSON 数组')
+    records = [header].concat(records)
+  }
+  const csvText = writeCSV(records, delim)
+  ctx.fs.writeFile(path, csvText)
+  return '已写入 CSV 文件 `' + path + '`（' + records.length + ' 行 × ' + records[0].length + ' 列，' + csvText.length + ' 字节）'
+}
+
+// json_to_table：JSON 数组 → Markdown 表格
+function jsonToTable(ctx, args) {
+  const jsonStr = String(args.json || '')
+  let records
+  try { records = JSON.parse(jsonStr) } catch { throw new Error('JSON 解析失败') }
+  if (!Array.isArray(records)) throw new Error('JSON 必须是数组')
+  if (records.length === 0) return '（空数组）'
+  const colSpec = String(args.columns || '').trim()
+  let limit = Number(args.limit || 100)
+  const title = String(args.title || '').trim()
+  let cols
+  if (colSpec) cols = colSpec.split(',').map(s => s.trim())
+  else cols = Object.keys(records[0] || {}).sort()
+  const total = records.length
+  if (limit > 0 && limit < records.length) records = records.slice(0, limit)
+  const rows = [cols]
+  for (const rec of records) {
+    rows.push(cols.map(col => (rec[col] !== undefined && rec[col] !== null) ? String(rec[col]) : ''))
+  }
+  let out = (title ? '**' + title + '**\n\n' : '') + '共 ' + total + ' 条记录'
+  if (limit > 0 && total > limit) out += '，显示前 ' + limit + ' 条'
+  out += '\n\n' + writeMarkdownTable(rows, null)
+  return out
+}
+
+// table_stats：表格数值列统计（求和/均值/最大/最小/计数，group_by 分组）
+function toRowsFromData(ctx, args, format, dataStr) {
+  if (format === 'file') {
+    const p = String(args.data || '').trim()
+    if (!p) throw new Error('data（文件路径）不能为空')
+    return readCSV(ctx.fs.readFile(p), ',')
+  }
+  if (format === 'json') {
+    const objs = JSON.parse(dataStr)
+    if (!Array.isArray(objs)) throw new Error('JSON 必须是数组')
+    if (objs.length === 0) return []
+    const cols = Object.keys(objs[0]).sort()
+    const rows = [cols]
+    for (const obj of objs) rows.push(cols.map(c => (obj[c] != null ? String(obj[c]) : '')))
+    return rows
+  }
+  return readCSV(dataStr, ',')
+}
+function writeStatsTable(stats) {
+  const rows = [['列名', '计数', '求和', '均值', '最小值', '最大值']]
+  for (const s of stats) {
+    const avg = s.count > 0 ? s.sum / s.count : 0
+    rows.push([s.name, String(s.count), s.sum.toFixed(2), avg.toFixed(2), s.min.toFixed(2), s.max.toFixed(2)])
+  }
+  return writeMarkdownTable(rows, null)
+}
+function tableStats(ctx, args) {
+  const dataStr = String(args.data || '')
+  const format = String(args.format || '').trim().toLowerCase() || 'csv'
+  const groupBy = String(args.group_by || '').trim()
+  const records = toRowsFromData(ctx, args, format, dataStr)
+  if (records.length < 2) return '（数据不足，至少需要表头 + 1 行数据）'
+  const header = records[0]
+  const dataRows = records.slice(1)
+  const colIdx = {}
+  header.forEach((h, i) => { colIdx[h] = i })
+  let groupColIdx = -1
+  if (groupBy && colIdx[groupBy] !== undefined) groupColIdx = colIdx[groupBy]
+  // 自动识别数值列
+  const numCols = []
+  if (dataRows.length > 0) {
+    header.forEach((h, ci) => {
+      if (ci === groupColIdx) return
+      let allNumeric = true
+      for (const row of dataRows) {
+        if (ci >= row.length) { allNumeric = false; break }
+        const v = String(row[ci]).trim()
+        if (v === '' || v === '-' || v === 'N/A') continue
+        if (isNaN(Number(v))) { allNumeric = false; break }
+      }
+      if (allNumeric) numCols.push({ name: h, sum: 0, count: 0, min: 0, max: 0 })
+    })
+  }
+  if (numCols.length === 0) return '（未找到数值列，无法统计）'
+  // 累计器
+  const accum = (st, row) => {
+    const ci = colIdx[st.name]
+    if (ci >= row.length) return
+    const v = String(row[ci]).trim()
+    if (v === '' || v === '-' || v === 'N/A' || isNaN(Number(v))) return
+    const f = Number(v)
+    st.sum += f
+    st.count++
+    if (st.count === 1) { st.min = f; st.max = f }
+    else { if (f < st.min) st.min = f; if (f > st.max) st.max = f }
+  }
+  if (groupColIdx >= 0) {
+    const groups = {}
+    dataRows.forEach(row => { const k = groupColIdx < row.length ? row[groupColIdx] : ''; (groups[k] || (groups[k] = [])).push(row) })
+    let out = '**分组统计**（按 `' + groupBy + '` 分组）\n\n'
+    Object.keys(groups).sort().forEach(gk => {
+      const stats = numCols.map(nc => {
+        const st = { name: nc.name, sum: 0, count: 0, min: 0, max: 0 }
+        groups[gk].forEach(row => accum(st, row))
+        return st
+      })
+      out += '**' + gk + '**（' + groups[gk].length + ' 行）\n\n' + writeStatsTable(stats) + '\n'
+    })
+    return out
+  }
+  let out = '**统计结果** · ' + dataRows.length + ' 行数据\n\n'
+  numCols.forEach(nc => { dataRows.forEach(row => accum(nc, row)) })
+  out += writeStatsTable(numCols)
+  return out
+}
+
+// text_report：目录树代码行数统计
+const SKIP_DIRS2 = { '.git': 1, 'node_modules': 1, 'vendor': 1, '.pair': 1, 'dist': 1, 'build': 1, 'target': 1, '.next': 1, '__pycache__': 1, 'coverage': 1, '.idea': 1, '.vscode': 1, '.cache': 1, '.venv': 1, 'venv': 1, 'Pods': 1 }
+function countFileLines(lines) {
+  let total = 0, code = 0, comment = 0, blank = 0
+  for (const line of lines) {
+    const t = line.trim()
+    total++
+    if (!t) blank++
+    else if (t.startsWith('//') || t.startsWith('#') || t.startsWith('--') || t.startsWith('/*') || t.startsWith('*') || t.startsWith('<!--')) comment++
+    else code++
+  }
+  return { total, code, comment, blank }
+}
+function textReport(ctx, args) {
+  const scanPath = String(args.path || '').trim() || '.'
+  const extStr = String(args.extensions || '').trim()
+  let groupBy = String(args.group_by || '').trim().toLowerCase() || 'ext'
+  const maxFiles = Number(args.max_files || 5000)
+  const extSet = {}
+  if (extStr) {
+    for (let e of extStr.split(',')) {
+      e = e.trim()
+      if (!e.startsWith('.')) e = '.' + e
+      extSet[e.toLowerCase()] = true
+    }
+  }
+  const extStats = {}, dirStats = {}
+  const total = { total: 0, code: 0, comment: 0, blank: 0 }
+  let fileCount = 0
+  const processFile = (p, rel) => {
+    if (fileCount >= maxFiles) return
+    let st
+    try { st = ctx.fs.stat(p) } catch { return }
+    if (st.isDir) return
+    const ext = p.toLowerCase().match(/(\.[a-z0-9]+)$/)?.[1] || '(无扩展名)'
+    if (Object.keys(extSet).length > 0 && !extSet[ext.toLowerCase()]) return
+    let data
+    try { data = ctx.fs.readFile(p) } catch { return }
+    fileCount++
+    const fs2 = countFileLines(String(data).split('\n'))
+    const dirKey = rel.includes('/') ? rel.slice(0, rel.lastIndexOf('/')) : '(根目录)'
+    if (!extStats[ext]) extStats[ext] = { total: 0, code: 0, comment: 0, blank: 0 }
+    const es = extStats[ext]
+    es.total += fs2.total; es.code += fs2.code; es.comment += fs2.comment; es.blank += fs2.blank
+    if (!dirStats[dirKey]) dirStats[dirKey] = { total: 0, code: 0, comment: 0, blank: 0 }
+    const ds = dirStats[dirKey]
+    ds.total += fs2.total; ds.code += fs2.code; ds.comment += fs2.comment; ds.blank += fs2.blank
+    total.total += fs2.total; total.code += fs2.code; total.comment += fs2.comment; total.blank += fs2.blank
+  }
+  const walk = (dir, relPrefix) => {
+    if (fileCount >= maxFiles) return
+    let names = []
+    try { names = ctx.fs.readdir(dir) } catch { return }
+    for (const n of names.sort()) {
+      if (fileCount >= maxFiles) return
+      if (SKIP_DIRS2[n]) continue
+      const p = (dir === '.' ? '' : dir + '/') + n
+      const rel = relPrefix ? relPrefix + '/' + n : n
+      let st
+      try { st = ctx.fs.stat(p) } catch { continue }
+      if (st.isDir) walk(p, rel)
+      else processFile(p, rel)
+    }
+  }
+  let scanSt
+  try { scanSt = ctx.fs.stat(scanPath) } catch { }
+  if (scanSt && !scanSt.isDir) processFile(scanPath, scanPath)
+  else walk(scanPath, '')
+  if (fileCount === 0) return '（未找到匹配的文件）'
+  let out = '**代码统计报告** · 扫描目录: `' + scanPath + '`\n\n共计 ' + fileCount + ' 个文件，' + total.total + ' 行（代码 ' + total.code + ' / 注释 ' + total.comment + ' / 空行 ' + total.blank + '）\n\n'
+  const rows = []
+  if (groupBy === 'dir') {
+    rows.push(['目录', '文件数（估算）', '总行数', '代码行', '注释行', '空行'])
+    Object.keys(dirStats).sort().forEach(d => {
+      const s = dirStats[d]
+      const estFiles = Math.ceil(s.total / 50)
+      rows.push([d, '~' + estFiles, String(s.total), String(s.code), String(s.comment), String(s.blank)])
+    })
+  } else {
+    rows.push(['扩展名', '文件数（估算）', '总行数', '代码行', '注释行', '空行'])
+    Object.keys(extStats).sort().forEach(e => {
+      const s = extStats[e]
+      const estFiles = Math.ceil(s.total / 50)
+      rows.push([e, '~' + estFiles, String(s.total), String(s.code), String(s.comment), String(s.blank)])
+    })
+  }
+  out += writeMarkdownTable(rows, null)
+  return out
+}
+
+const impls = {
+  csv_read: csvRead,
+  csv_write: csvWrite,
+  json_to_table: jsonToTable,
+  table_stats: tableStats,
+  text_report: textReport,
+  word_read: wordRead,
+  read_xlsx: readXlsx,
+  // word_write / write_xlsx / read_pdf / markdown_to_html 保留独立二进制（生成/PDF 引擎）
+}
+
+// ── read_xlsx 辅助 ─────────────────────────────────────────
+// Excel 列字母 → 索引（A→1, Z→26, AA→27）
+function parseColLetter(letters) {
+  if (!letters) return 0
+  let col = 0
+  for (const ch of String(letters).toUpperCase()) col = col * 26 + (ch.charCodeAt(0) - 64)
+  return col
+}
+// 剥离 XML 命名空间声明（对齐 Go 版 clean）
+function cfgStripXmlns(s) { return String(s || '').replace(/xmlns="[^"]*"/g, '') }
+
+// ── word_read：.docx 纯文本/Markdown 提取（zipReadEntry → JS XML 解析）─────
+function wordRead(ctx, args) {
+  const path = String(args.path || '')
+  const format = (String(args.format || 'text') || 'text').toLowerCase() === 'markdown' ? 'markdown' : 'text'
+  const limit = parseInt(args.limit, 10)
+  const lim = isNaN(limit) ? 10000 : limit
+  let xml
+  try { xml = ctx.fs.zipReadEntry(path, 'word/document.xml') } catch (e) { throw new Error('无法打开 .docx 文件（不是有效的 ZIP 压缩包）: ' + (e && e.message || e)) }
+  if (!xml || !xml.includes('document')) xml = String(xml || '')
+  // 先剥离表格块（表格单元格内的段落不算文档段落）
+  const tbls = []
+  let body = String(xml).replace(/<w:tbl>[\s\S]*?<\/w:tbl>/g, (m) => { tbls.push(m); return '\n@@TBL@@\n' })
+  // 提取段落
+  const paras = []
+  const reP = /<w:p\b[^>]*>([\s\S]*?)<\/w:p>/g
+  let m
+  while ((m = reP.exec(body)) !== null) {
+    const inner = m[1]
+    const styleM = inner.match(/<w:pStyle[^>]*w:val="([^"]+)"[^>]*\/?>/)
+    const style = styleM ? styleM[1] : ''
+    let text = ''
+    const reT = /<w:t[^>]*>([^<]*)<\/w:t>/g
+    let tm
+    while ((tm = reT.exec(inner)) !== null) text += tm[1]
+    paras.push({ text: text, style: style })
+  }
+  // 输出
+  let out = ''
+  for (const p of paras) {
+    if (!p.text) { out += format === 'markdown' ? '\n\n' : '\n'; continue }
+    if (format === 'markdown') {
+      if (/^(Heading1|1)/.test(p.style)) out += '# ' + p.text + '\n\n'
+      else if (/^(Heading2|2)/.test(p.style)) out += '## ' + p.text + '\n\n'
+      else if (/^(Heading3|3)/.test(p.style)) out += '### ' + p.text + '\n\n'
+      else if (/ListBullet/.test(p.style)) out += '- ' + p.text + '\n'
+      else if (/ListNumber/.test(p.style)) out += '1. ' + p.text + '\n'
+      else out += p.text + '\n\n'
+    } else {
+      out += p.text + '\n'
+    }
+  }
+  // 表格
+  for (const tb of tbls) {
+    if (format === 'markdown') {
+      const rows = []
+      const reTr = /<w:tr[^>]*>([\s\S]*?)<\/w:tr>/g
+      let tm, tr
+      while ((tm = reTr.exec(tb)) !== null) {
+        const cells = []
+        const reTc = /<w:tc[^>]*>([\s\S]*?)<\/w:tc>/g
+        let cm, tc
+        while ((cm = reTc.exec(tm[1])) !== null) {
+          let ct = ''
+          const reCt = /<w:t[^>]*>([^<]*)<\/w:t>/g
+          let xt
+          while ((xt = reCt.exec(cm[1])) !== null) ct += xt[1]
+          cells.push(ct.trim())
+        }
+        if (cells.length) rows.push(cells)
+      }
+      if (rows.length) { out += '\n' + writeMarkdownTable(rows, null) + '\n' }
+    } else {
+      out += '\n[表格]\n'
+      const reTr = /<w:tr[^>]*>([\s\S]*?)<\/w:tr>/g
+      let tm, tr
+      while ((tm = reTr.exec(tb)) !== null) {
+        const cells = []
+        const reTc = /<w:tc[^>]*>([\s\S]*?)<\/w:tc>/g
+        let cm, tc
+        while ((cm = reTc.exec(tm[1])) !== null) {
+          let ct = ''
+          const reCt = /<w:t[^>]*>([^<]*)<\/w:t>/g
+          let xt
+          while ((xt = reCt.exec(cm[1])) !== null) ct += xt[1]
+          cells.push(ct.trim())
+        }
+        if (cells.length) out += '| ' + cells.join(' | ') + ' |\n'
+      }
+      out += '\n'
+    }
+  }
+  let result = out.trim()
+  if (lim > 0 && result.length > lim) result = result.slice(0, lim) + '…'
+  return result
+}
+
+// ── read_xlsx：.xlsx → Markdown 表格（zip 条目 → JS XML 解析）─────────────
+function readXlsx(ctx, args) {
+  const path = String(args.path || '')
+  const sheetName = String(args.sheet || '').trim()
+  const limit = parseInt(args.limit, 10)
+  const lim = isNaN(limit) ? 200 : limit
+  // 共享字符串表
+  let sharedStrings = []
+  try {
+    const ssx = ctx.fs.zipReadEntry(path, 'xl/sharedStrings.xml')
+    const siRe = /<si>([\s\S]*?)<\/si>/g
+    let m
+    while ((m = siRe.exec(ssx)) !== null) {
+      let t = ''
+      const tRe = /<t[^>]*>([^<]*)<\/t>/g
+      let tm
+      while ((tm = tRe.exec(m[1])) !== null) t += tm[1]
+      sharedStrings.push(t)
+    }
+  } catch (e) { /* 无共享字符串表 */ }
+  // 找第一个 sheet（★ 与 Go 版一致：跳过 workbook 解析，取 sheet*.xml 首个）
+  const entries = ctx.fs.zipEntries(path)
+  const sheets = entries.filter(n => n.indexOf('xl/worksheets/sheet') === 0 && n.endsWith('.xml')).sort()
+  if (sheets.length === 0) return '（空文件或未找到数据）'
+  const sx = cfgStripXmlns(ctx.fs.zipReadEntry(path, sheets[0]))
+  // 行 → cell
+  const rows = []
+  const reRow = /<row[^>]*>([\s\S]*?)<\/row>/g
+  let m2
+  while ((m2 = reRow.exec(sx)) !== null) {
+    const cells = []
+    const reC = /<c\b([^>]*)>([\s\S]*?)<\/c>|<c\b([^>]*)\/>/g
+    let cm
+    while ((cm = reC.exec(m2[1])) !== null) {
+      const attrs = cm[1] || cm[3] || ''
+      const inner = cm[2] || ''
+      const refM = attrs.match(/r="([A-Za-z]+)\d+"/)
+      const typeM = attrs.match(/t="([^"]+)"/)
+      const colLetter = refM ? refM[1] : ''
+      let col = parseColLetter(colLetter)
+      if (col === 0) col = cells.length + 1
+      let val = ''
+      const vM = inner.match(/<v>([^<]*)<\/v>/)
+      if (vM) val = vM[1]
+      const isM = inner.match(/<is>([\s\S]*?)<\/is>/)
+      if (!val && isM) {
+        let t = ''
+        const tRe = /<t[^>]*>([^<]*)<\/t>/g
+        let tm
+        while ((tm = tRe.exec(isM[1])) !== null) t += tm[1]
+        val = t
+      }
+      const type = typeM ? typeM[1] : ''
+      if (type === 's') {
+        const idx = parseInt(val, 10)
+        val = (!isNaN(idx) && idx >= 0 && idx < sharedStrings.length) ? sharedStrings[idx] : val
+      }
+      cells.push({ col: col, val: val })
+    }
+    if (cells.length) rows.push(cells)
+  }
+  if (rows.length === 0) return '（空文件或未找到数据）'
+  // 最大列数
+  let maxCol = 0
+  for (const row of rows) for (const c of row.cells || row) maxCol = Math.max(maxCol, (c && c.col) || 0)
+  if (maxCol === 0) { for (const row of rows) maxCol = Math.max(maxCol, row.length) }
+  // 构造记录
+  let shown = rows
+  if (lim > 0 && lim < rows.length) shown = rows.slice(0, lim)
+  const records = []
+  for (const row of shown) {
+    const rec = new Array(maxCol).fill('')
+    for (const c of (row.cells || row)) {
+      const idx = (c && c.col || 0) - 1
+      if (idx >= 0 && idx < maxCol) rec[idx] = c.val
+    }
+    records.push(rec)
+  }
+  let out = `工作表: **${sheetName}** · 共 ${rows.length} 行 × ${maxCol} 列`
+  if (lim > 0 && rows.length > lim) out += ` · 显示前 ${lim} 行`
+  out += '\n\n' + writeMarkdownTable(records, null)
+  return out
+}
+
+
 return {
   name: 'tool-office',
-  purpose: '办公文档（csv_read/csv_write/json_to_table/table_stats/text_report/word_read）（自动生成，迁移自内置 Go 工具组）',
+  inject: ['fs'],
+  purpose: '办公文档（csv_read/csv_write/json_to_table/table_stats/text_report/word_read/read_xlsx JS 原生化；word_write/write_xlsx/read_pdf/markdown_to_html 保留独立二进制）——2026-08-22 迁移',
   apply(ctx) {
     for (const t of tools) {
       ctx.tools.register({
@@ -328,7 +885,7 @@ return {
         requiresApproval: t.requiresApproval,
         systemTool: t.systemTool,
         parameters: t.parameters,
-        execute: (args) => ctx.binary.exec(t.name, args || {}),
+        execute: (args) => (impls[t.name] ? impls[t.name](ctx, args || {}) : ctx.binary.exec(t.name, args || {})),
       })
     }
   },

package/package.json

@@ -3,7 +3,7 @@
   "name": "@paircode/tool-office",
   "scope": "global",
   "type": "plugin",
-  "version": "1.0.1",
+  "version": "1.0.3",
   "description": "办公文档（csv_read/csv_write/json_to_table/table_stats/text_report/word_read）（自动生成，迁移自内置 Go 工具组）",
   "keywords": [
     "paircode"