kordoc 0.1.1 → 0.2.2

package/dist/{chunk-P2BZDRLZ.js → chunk-KZMWHK72.js}

@@ -1,28 +1,20 @@
 #!/usr/bin/env node
-var __defProp = Object.defineProperty;
-var __require = /* @__PURE__ */ ((x) => typeof require !== "undefined" ? require : typeof Proxy !== "undefined" ? new Proxy(x, {
-  get: (a, b) => (typeof require !== "undefined" ? require : a)[b]
-}) : x)(function(x) {
-  if (typeof require !== "undefined") return require.apply(this, arguments);
-  throw Error('Dynamic require of "' + x + '" is not supported');
-});
-var __export = (target, all) => {
-  for (var name in all)
-    __defProp(target, name, { get: all[name], enumerable: true });
-};
 
 // src/detect.ts
+function magicBytes(buffer) {
+  return new Uint8Array(buffer, 0, Math.min(4, buffer.byteLength));
+}
 function isHwpxFile(buffer) {
-  const bytes = new Uint8Array(buffer.slice(0, 4));
-  return bytes[0] === 80 && bytes[1] === 75 && bytes[2] === 3 && bytes[3] === 4;
+  const b = magicBytes(buffer);
+  return b[0] === 80 && b[1] === 75 && b[2] === 3 && b[3] === 4;
 }
 function isOldHwpFile(buffer) {
-  const bytes = new Uint8Array(buffer.slice(0, 4));
-  return bytes[0] === 208 && bytes[1] === 207 && bytes[2] === 17 && bytes[3] === 224;
+  const b = magicBytes(buffer);
+  return b[0] === 208 && b[1] === 207 && b[2] === 17 && b[3] === 224;
 }
 function isPdfFile(buffer) {
-  const bytes = new Uint8Array(buffer.slice(0, 4));
-  return bytes[0] === 37 && bytes[1] === 80 && bytes[2] === 68 && bytes[3] === 70;
+  const b = magicBytes(buffer);
+  return b[0] === 37 && b[1] === 80 && b[2] === 68 && b[3] === 70;
 }
 function detectFormat(buffer) {
   if (isHwpxFile(buffer)) return "hwpx";
@@ -31,29 +23,40 @@ function detectFormat(buffer) {
   return "unknown";
 }
 
+// src/utils.ts
+var VERSION = true ? "0.2.2" : "0.0.0-dev";
+function toArrayBuffer(buf) {
+  return buf.buffer.slice(buf.byteOffset, buf.byteOffset + buf.byteLength);
+}
+
 // src/hwpx/parser.ts
 import JSZip from "jszip";
+import { inflateRawSync } from "zlib";
 import { DOMParser } from "@xmldom/xmldom";
 
 // src/table/builder.ts
+var MAX_COLS = 200;
+var MAX_ROWS = 1e4;
 function buildTable(rows) {
+  if (rows.length > MAX_ROWS) rows = rows.slice(0, MAX_ROWS);
   const numRows = rows.length;
-  const tempOccupied = Array.from({ length: numRows }, () => Array(100).fill(false));
+  const tempOccupied = /* @__PURE__ */ new Set();
   let maxCols = 0;
   for (let rowIdx = 0; rowIdx < numRows; rowIdx++) {
     let colIdx = 0;
     for (const cell of rows[rowIdx]) {
-      while (colIdx < 100 && tempOccupied[rowIdx][colIdx]) colIdx++;
-      if (colIdx >= 100) break;
+      while (colIdx < MAX_COLS && tempOccupied.has(rowIdx * MAX_COLS + colIdx)) colIdx++;
+      if (colIdx >= MAX_COLS) break;
       for (let r = rowIdx; r < Math.min(rowIdx + cell.rowSpan, numRows); r++) {
-        for (let c = colIdx; c < Math.min(colIdx + cell.colSpan, 100); c++) {
-          tempOccupied[r][c] = true;
+        for (let c = colIdx; c < Math.min(colIdx + cell.colSpan, MAX_COLS); c++) {
+          tempOccupied.add(r * MAX_COLS + c);
         }
       }
       colIdx += cell.colSpan;
       if (colIdx > maxCols) maxCols = colIdx;
     }
   }
+  tempOccupied.clear();
   if (maxCols === 0) return { rows: 0, cols: 0, cells: [], hasHeader: false };
   const grid = Array.from(
     { length: numRows },
@@ -146,13 +149,9 @@ function tableToMarkdown(table) {
     }
   }
   const uniqueRows = [];
-  const seen = /* @__PURE__ */ new Set();
   for (const row of display) {
-    const key = row.join("||");
-    if (!seen.has(key)) {
-      seen.add(key);
-      uniqueRows.push(row);
-    }
+    const isEmptyPlaceholder = row.every((cell) => cell === "");
+    if (!isEmptyPlaceholder) uniqueRows.push(row);
   }
   if (uniqueRows.length === 0) return "";
   const md = [];
@@ -165,6 +164,14 @@ function tableToMarkdown(table) {
 }
 
 // src/hwpx/parser.ts
+var MAX_DECOMPRESS_SIZE = 100 * 1024 * 1024;
+var MAX_ZIP_ENTRIES = 500;
+function clampSpan(val, max) {
+  return Math.max(1, Math.min(val, max));
+}
+function stripDtd(xml) {
+  return xml.replace(/<!DOCTYPE\s[^[>]*(\[[\s\S]*?\])?\s*>/gi, "");
+}
 async function parseHwpxDocument(buffer) {
   let zip;
   try {
@@ -174,11 +181,14 @@ async function parseHwpxDocument(buffer) {
   }
   const sectionPaths = await resolveSectionPaths(zip);
   if (sectionPaths.length === 0) throw new Error("HWPX\uC5D0\uC11C \uC139\uC158 \uD30C\uC77C\uC744 \uCC3E\uC744 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4");
+  let totalDecompressed = 0;
   const blocks = [];
   for (const path of sectionPaths) {
     const file = zip.file(path);
     if (!file) continue;
     const xml = await file.async("text");
+    totalDecompressed += xml.length * 2;
+    if (totalDecompressed > MAX_DECOMPRESS_SIZE) throw new Error("ZIP \uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
     blocks.push(...parseSectionXml(xml));
   }
   return blocksToMarkdown(blocks);
@@ -188,15 +198,31 @@ function extractFromBrokenZip(buffer) {
   const view = new DataView(buffer);
   let pos = 0;
   const texts = [];
+  let totalDecompressed = 0;
+  let entryCount = 0;
   while (pos < data.length - 30) {
     if (data[pos] !== 80 || data[pos + 1] !== 75 || data[pos + 2] !== 3 || data[pos + 3] !== 4) break;
+    if (++entryCount > MAX_ZIP_ENTRIES) break;
     const method = view.getUint16(pos + 8, true);
     const compSize = view.getUint32(pos + 18, true);
     const nameLen = view.getUint16(pos + 26, true);
     const extraLen = view.getUint16(pos + 28, true);
+    if (nameLen > 1024 || extraLen > 65535) {
+      pos += 30 + nameLen + extraLen;
+      continue;
+    }
+    const fileStart = pos + 30 + nameLen + extraLen;
+    if (fileStart + compSize > data.length) break;
+    if (compSize === 0 && method !== 0) {
+      pos = fileStart;
+      continue;
+    }
     const nameBytes = data.slice(pos + 30, pos + 30 + nameLen);
     const name = new TextDecoder().decode(nameBytes);
-    const fileStart = pos + 30 + nameLen + extraLen;
+    if (name.includes("..") || name.startsWith("/")) {
+      pos = fileStart + compSize;
+      continue;
+    }
     const fileData = data.slice(fileStart, fileStart + compSize);
     pos = fileStart + compSize;
     if (!name.toLowerCase().includes("section") || !name.endsWith(".xml")) continue;
@@ -205,11 +231,13 @@ function extractFromBrokenZip(buffer) {
       if (method === 0) {
         content = new TextDecoder().decode(fileData);
       } else if (method === 8) {
-        const { inflateRawSync: inflateRawSync2 } = __require("zlib");
-        content = new TextDecoder().decode(inflateRawSync2(Buffer.from(fileData)));
+        const decompressed = inflateRawSync(Buffer.from(fileData), { maxOutputLength: MAX_DECOMPRESS_SIZE });
+        content = new TextDecoder().decode(decompressed);
       } else {
         continue;
       }
+      totalDecompressed += content.length * 2;
+      if (totalDecompressed > MAX_DECOMPRESS_SIZE) throw new Error("\uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC");
       const sectionText = blocksToMarkdown(parseSectionXml(content));
       if (sectionText) texts.push(sectionText);
     } catch {
@@ -233,7 +261,7 @@ async function resolveSectionPaths(zip) {
 }
 function parseSectionPathsFromManifest(xml) {
   const parser = new DOMParser();
-  const doc = parser.parseFromString(xml, "text/xml");
+  const doc = parser.parseFromString(stripDtd(xml), "text/xml");
   const items = doc.getElementsByTagName("opf:item");
   const spine = doc.getElementsByTagName("opf:itemref");
   const isSectionId = (id) => /^s/i.test(id) || id.toLowerCase().includes("section");
@@ -260,7 +288,7 @@ function parseSectionPathsFromManifest(xml) {
 }
 function parseSectionXml(xml) {
   const parser = new DOMParser();
-  const doc = parser.parseFromString(xml, "text/xml");
+  const doc = parser.parseFromString(stripDtd(xml), "text/xml");
   if (!doc.documentElement) return [];
   const blocks = [];
   walkSection(doc.documentElement, blocks, null, []);
@@ -318,8 +346,8 @@ function walkSection(node, blocks, tableCtx, tableStack) {
         if (tableCtx?.cell) {
           const cs = parseInt(el.getAttribute("colSpan") || "1", 10);
           const rs = parseInt(el.getAttribute("rowSpan") || "1", 10);
-          if (cs > 0) tableCtx.cell.colSpan = cs;
-          if (rs > 0) tableCtx.cell.rowSpan = rs;
+          tableCtx.cell.colSpan = clampSpan(cs, MAX_COLS);
+          tableCtx.cell.rowSpan = clampSpan(rs, MAX_ROWS);
         }
         break;
       case "p": {
@@ -381,7 +409,7 @@ function extractParagraphText(para) {
 }
 
 // src/hwp5/record.ts
-import { inflateRawSync, inflateSync } from "zlib";
+import { inflateRawSync as inflateRawSync2, inflateSync } from "zlib";
 var TAG_PARA_HEADER = 66;
 var TAG_PARA_TEXT = 67;
 var TAG_CTRL_HEADER = 71;
@@ -416,16 +444,19 @@ function readRecords(data) {
   }
   return records;
 }
+var MAX_DECOMPRESS_SIZE2 = 100 * 1024 * 1024;
 function decompressStream(data) {
+  const opts = { maxOutputLength: MAX_DECOMPRESS_SIZE2 };
   if (data.length >= 2 && data[0] === 120) {
     try {
-      return inflateSync(data);
+      return inflateSync(data, opts);
     } catch {
     }
   }
-  return inflateRawSync(data);
+  return inflateRawSync2(data, opts);
 }
 function parseFileHeader(data) {
+  if (data.length < 40) throw new Error("FileHeader\uAC00 \uB108\uBB34 \uC9E7\uC2B5\uB2C8\uB2E4 (\uCD5C\uC18C 40\uBC14\uC774\uD2B8)");
   const sig = data.subarray(0, 32).toString("utf8").replace(/\0+$/, "");
   return {
     signature: sig,
@@ -461,6 +492,15 @@ function extractText(data) {
           const isInline = ch >= 4 && ch <= 9 || ch >= 19 && ch <= 20;
           if ((isExt || isInline) && i + 14 <= data.length) i += 14;
         } else if (ch >= 32) {
+          if (ch >= 55296 && ch <= 56319 && i + 1 < data.length) {
+            const lo = data.readUInt16LE(i);
+            if (lo >= 56320 && lo <= 57343) {
+              i += 2;
+              const codePoint = (ch - 55296 << 10) + (lo - 56320) + 65536;
+              result += String.fromCodePoint(codePoint);
+              break;
+            }
+          }
           result += String.fromCharCode(ch);
         }
         break;
@@ -473,6 +513,8 @@ function extractText(data) {
 import { createRequire } from "module";
 var require2 = createRequire(import.meta.url);
 var CFB = require2("cfb");
+var MAX_SECTIONS = 100;
+var MAX_TOTAL_DECOMPRESS = 100 * 1024 * 1024;
 function parseHwp5Document(buffer) {
   const cfb = CFB.parse(buffer);
   const headerEntry = CFB.find(cfb, "/FileHeader");
@@ -485,8 +527,11 @@ function parseHwp5Document(buffer) {
   const sections = findSections(cfb);
   if (sections.length === 0) throw new Error("\uC139\uC158 \uC2A4\uD2B8\uB9BC\uC744 \uCC3E\uC744 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4");
   const blocks = [];
+  let totalDecompressed = 0;
   for (const sectionData of sections) {
     const data = compressed ? decompressStream(Buffer.from(sectionData)) : Buffer.from(sectionData);
+    totalDecompressed += data.length;
+    if (totalDecompressed > MAX_TOTAL_DECOMPRESS) throw new Error("\uCD1D \uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC (decompression bomb \uC758\uC2EC)");
     const records = readRecords(data);
     blocks.push(...parseSection(records));
   }
@@ -494,7 +539,7 @@ function parseHwp5Document(buffer) {
 }
 function findSections(cfb) {
   const sections = [];
-  for (let i = 0; ; i++) {
+  for (let i = 0; i < MAX_SECTIONS; i++) {
     const entry = CFB.find(cfb, `/BodyText/Section${i}`);
     if (!entry?.content) break;
     sections.push({ idx: i, content: Buffer.from(entry.content) });
@@ -585,20 +630,29 @@ function parseTableBlock(records, startIdx) {
   return { table: buildTable(cellRows), nextIdx: i };
 }
 function parseCellBlock(records, startIdx, tableLevel) {
-  const cellLevel = records[startIdx].level;
+  const rec = records[startIdx];
+  const cellLevel = rec.level;
   const texts = [];
+  let colSpan = 1;
+  let rowSpan = 1;
+  if (rec.data.length >= 14) {
+    const cs = rec.data.readUInt16LE(10);
+    const rs = rec.data.readUInt16LE(12);
+    if (cs > 0) colSpan = Math.min(cs, MAX_COLS);
+    if (rs > 0) rowSpan = Math.min(rs, MAX_ROWS);
+  }
   let i = startIdx + 1;
   while (i < records.length) {
-    const rec = records[i];
-    if (rec.tagId === TAG_LIST_HEADER && rec.level <= cellLevel) break;
-    if (rec.level <= tableLevel && (rec.tagId === TAG_PARA_HEADER || rec.tagId === TAG_CTRL_HEADER)) break;
-    if (rec.tagId === TAG_PARA_TEXT) {
-      const t = extractText(rec.data).trim();
+    const r = records[i];
+    if (r.tagId === TAG_LIST_HEADER && r.level <= cellLevel) break;
+    if (r.level <= tableLevel && (r.tagId === TAG_PARA_HEADER || r.tagId === TAG_CTRL_HEADER)) break;
+    if (r.tagId === TAG_PARA_TEXT) {
+      const t = extractText(r.data).trim();
       if (t) texts.push(t);
     }
     i++;
   }
-  return { cell: { text: texts.join("\n"), colSpan: 1, rowSpan: 1 }, nextIdx: i };
+  return { cell: { text: texts.join("\n"), colSpan, rowSpan }, nextIdx: i };
 }
 function arrangeCells(rows, cols, cells) {
   const grid = Array.from({ length: rows }, () => Array(cols).fill(null));
@@ -627,13 +681,18 @@ var pdfjsModule = null;
 async function loadPdfjs() {
   if (pdfjsModule) return pdfjsModule;
   try {
-    pdfjsModule = await import("pdfjs-dist/legacy/build/pdf.mjs");
+    const mod = await import("pdfjs-dist/legacy/build/pdf.mjs");
     const req = createRequire2(import.meta.url);
     const workerPath = req.resolve("pdfjs-dist/legacy/build/pdf.worker.mjs");
-    pdfjsModule.GlobalWorkerOptions.workerSrc = pathToFileURL(workerPath).href;
-    return pdfjsModule;
-  } catch {
-    return null;
+    mod.GlobalWorkerOptions.workerSrc = pathToFileURL(workerPath).href;
+    pdfjsModule = mod;
+    return mod;
+  } catch (err) {
+    const msg = err instanceof Error ? err.message : String(err);
+    if (msg.includes("Cannot find") || msg.includes("MODULE_NOT_FOUND")) {
+      return null;
+    }
+    throw new Error(`pdfjs-dist \uB85C\uB529 \uC2E4\uD328: ${msg}`);
   }
 }
 async function parsePdfDocument(buffer) {
@@ -646,15 +705,14 @@ async function parsePdfDocument(buffer) {
       error: "pdfjs-dist\uAC00 \uC124\uCE58\uB418\uC9C0 \uC54A\uC558\uC2B5\uB2C8\uB2E4. npm install pdfjs-dist"
     };
   }
+  const data = new Uint8Array(buffer);
+  const doc = await pdfjs.getDocument({
+    data,
+    useSystemFonts: true,
+    disableFontFace: true,
+    isEvalSupported: false
+  }).promise;
   try {
-    const data = new Uint8Array(buffer);
-    const doc = await pdfjs.getDocument({
-      data,
-      useSystemFonts: true,
-      disableFontFace: true,
-      isEvalSupported: false,
-      workerSrc: ""
-    }).promise;
     const pageCount = doc.numPages;
     if (pageCount === 0) {
       return { success: false, fileType: "pdf", pageCount: 0, error: "PDF\uC5D0 \uD398\uC774\uC9C0\uAC00 \uC5C6\uC2B5\uB2C8\uB2E4." };
@@ -689,20 +747,14 @@ async function parsePdfDocument(buffer) {
     }
     markdown = reconstructTables(markdown);
     return { success: true, fileType: "pdf", markdown, pageCount, isImageBased: false };
-  } catch (err) {
-    return {
-      success: false,
-      fileType: "pdf",
-      pageCount: 0,
-      error: err instanceof Error ? err.message : "PDF \uD30C\uC2F1 \uC2E4\uD328"
-    };
+  } finally {
+    await doc.destroy().catch(() => {
+    });
   }
 }
 function groupTextItemsByLine(items) {
   if (items.length === 0) return [];
-  const textItems = items.filter(
-    (item) => typeof item.str === "string" && item.str.trim() !== ""
-  );
+  const textItems = items.filter((item) => typeof item.str === "string" && item.str.trim() !== "");
   if (textItems.length === 0) return [];
   textItems.sort((a, b) => {
     const yDiff = b.transform[5] - a.transform[5];
@@ -760,8 +812,9 @@ function reconstructTables(text) {
 function formatAsMarkdownTable(rows) {
   const maxCols = Math.max(...rows.map((r) => r.length));
   const normalized = rows.map((r) => {
-    while (r.length < maxCols) r.push("");
-    return r;
+    const copy = [...r];
+    while (copy.length < maxCols) copy.push("");
+    return copy;
   });
   const lines = [];
   lines.push("| " + normalized[0].join(" | ") + " |");
@@ -774,6 +827,9 @@ function formatAsMarkdownTable(rows) {
 
 // src/index.ts
 async function parse(buffer) {
+  if (!buffer || buffer.byteLength === 0) {
+    return { success: false, fileType: "unknown", error: "\uBE48 \uBC84\uD37C\uC774\uAC70\uB098 \uC720\uD6A8\uD558\uC9C0 \uC54A\uC740 \uC785\uB825\uC785\uB2C8\uB2E4." };
+  }
   const format = detectFormat(buffer);
   switch (format) {
     case "hwpx":
@@ -803,11 +859,16 @@ async function parseHwp(buffer) {
   }
 }
 async function parsePdf(buffer) {
-  return parsePdfDocument(buffer);
+  try {
+    return await parsePdfDocument(buffer);
+  } catch (err) {
+    return { success: false, fileType: "pdf", error: err instanceof Error ? err.message : "PDF \uD30C\uC2F1 \uC2E4\uD328" };
+  }
 }
 
 export {
-  __export,
   detectFormat,
+  VERSION,
+  toArrayBuffer,
   parse
 };

package/dist/cli.js

@@ -1,21 +1,31 @@
 #!/usr/bin/env node
 import {
+  VERSION,
   detectFormat,
-  parse
-} from "./chunk-P2BZDRLZ.js";
+  parse,
+  toArrayBuffer
+} from "./chunk-KZMWHK72.js";
 
 // src/cli.ts
-import { readFileSync, writeFileSync } from "fs";
+import { readFileSync, writeFileSync, mkdirSync, statSync } from "fs";
 import { basename, resolve } from "path";
 import { Command } from "commander";
 var program = new Command();
-program.name("kordoc").description("\uBAA8\uB450 \uD30C\uC2F1\uD574\uBC84\uB9AC\uACA0\uB2E4 \u2014 HWP, HWPX, PDF \u2192 Markdown").version("0.1.0").argument("<files...>", "\uBCC0\uD658\uD560 \uD30C\uC77C \uACBD\uB85C (HWP, HWPX, PDF)").option("-o, --output <path>", "\uCD9C\uB825 \uD30C\uC77C \uACBD\uB85C (\uB2E8\uC77C \uD30C\uC77C \uC2DC)").option("-d, --out-dir <dir>", "\uCD9C\uB825 \uB514\uB809\uD1A0\uB9AC (\uB2E4\uC911 \uD30C\uC77C \uC2DC)").option("--format <type>", "\uCD9C\uB825 \uD615\uC2DD: markdown (\uAE30\uBCF8) \uB610\uB294 json", "markdown").option("--silent", "\uC9C4\uD589 \uBA54\uC2DC\uC9C0 \uC228\uAE30\uAE30").action(async (files, opts) => {
+program.name("kordoc").description("\uBAA8\uB450 \uD30C\uC2F1\uD574\uBC84\uB9AC\uACA0\uB2E4 \u2014 HWP, HWPX, PDF \u2192 Markdown").version(VERSION).argument("<files...>", "\uBCC0\uD658\uD560 \uD30C\uC77C \uACBD\uB85C (HWP, HWPX, PDF)").option("-o, --output <path>", "\uCD9C\uB825 \uD30C\uC77C \uACBD\uB85C (\uB2E8\uC77C \uD30C\uC77C \uC2DC)").option("-d, --out-dir <dir>", "\uCD9C\uB825 \uB514\uB809\uD1A0\uB9AC (\uB2E4\uC911 \uD30C\uC77C \uC2DC)").option("--format <type>", "\uCD9C\uB825 \uD615\uC2DD: markdown (\uAE30\uBCF8) \uB610\uB294 json", "markdown").option("--silent", "\uC9C4\uD589 \uBA54\uC2DC\uC9C0 \uC228\uAE30\uAE30").action(async (files, opts) => {
   for (const filePath of files) {
     const absPath = resolve(filePath);
     const fileName = basename(absPath);
     try {
+      const fileSize = statSync(absPath).size;
+      if (fileSize > 500 * 1024 * 1024) {
+        process.stderr.write(`
+[kordoc] SKIP: ${fileName} \u2014 \uD30C\uC77C\uC774 \uB108\uBB34 \uD07D\uB2C8\uB2E4 (${(fileSize / 1024 / 1024).toFixed(1)}MB)
+`);
+        process.exitCode = 1;
+        continue;
+      }
       const buffer = readFileSync(absPath);
-      const arrayBuffer = buffer.buffer.slice(buffer.byteOffset, buffer.byteOffset + buffer.byteLength);
+      const arrayBuffer = toArrayBuffer(buffer);
       const format = detectFormat(arrayBuffer);
       if (!opts.silent) {
         process.stderr.write(`[kordoc] ${fileName} (${format}) ...`);
@@ -31,13 +41,15 @@ program.name("kordoc").description("\uBAA8\uB450 \uD30C\uC2F1\uD574\uBC84\uB9AC\
       }
       if (!opts.silent) process.stderr.write(` OK
 `);
-      const output = opts.format === "json" ? JSON.stringify(result, null, 2) : result.markdown || "";
+      const output = opts.format === "json" ? JSON.stringify(result, null, 2) : result.markdown;
       if (opts.output && files.length === 1) {
         writeFileSync(opts.output, output, "utf-8");
         if (!opts.silent) process.stderr.write(`  \u2192 ${opts.output}
 `);
       } else if (opts.outDir) {
-        const outPath = resolve(opts.outDir, fileName.replace(/\.[^.]+$/, ".md"));
+        mkdirSync(opts.outDir, { recursive: true });
+        const outExt = opts.format === "json" ? ".json" : ".md";
+        const outPath = resolve(opts.outDir, fileName.replace(/\.[^.]+$/, outExt));
         writeFileSync(outPath, output, "utf-8");
         if (!opts.silent) process.stderr.write(`  \u2192 ${outPath}
 `);