@@ -12,6 +12,7 @@ import {
   isHtmlComplexityError,
 } from '@/lib/file-parsers/html-parser'
 import { parseOfficeText } from '@/lib/file-parsers/officeparser-module'
+import { isEncryptedOoxmlContainer } from '@/lib/file-parsers/ooxml-encryption'
 import type { FileParseOptions, FileParseResult, FileParser } from '@/lib/file-parsers/types'
 import { sanitizeTextForUTF8 } from '@/lib/file-parsers/utils'
 import { assertOoxmlArchiveWithinLimits } from '@/lib/file-parsers/zip-guard'
@@ -106,6 +107,14 @@ export class DocxParser implements FileParser {
         extractionErrors.push(officeError)
       }
 
+      if (isEncryptedOoxmlContainer(buffer)) {
+        throw new FileParserError(
+          'encrypted_file',
+          'This document is encrypted or password-protected',
+          extractionErrors.length > 0 ? new AggregateError(extractionErrors) : undefined
+        )
+      }
+
       const isZipFile = buffer.length >= 2 && buffer[0] === 0x50 && buffer[1] === 0x4b
       if (!isZipFile) {
         const textContent = buffer.toString('utf8').trim()
 
@@ -141,6 +141,46 @@ describe('HtmlParser', () => {
       expect(result.content).not.toContain('fourth nested')
     })
 
+    it('renders a nested table inside its cell exactly once', async () => {
+      const buffer = Buffer.from(
+        `<body><table><tbody><tr><td>Outer A</td><td><p>Intro</p>` +
+          `<table><tr><td>Inner 1</td><td>Inner 2</td></tr><tr><td>Inner 3</td></tr></table>` +
+          `</td></tr><tr><th>Outer B</th><td>Plain</td></tr></tbody></table></body>`
+      )
+
+      const result = await parser.parseBuffer(buffer)
+
+      expect(result.content).toContain('| Outer A | Intro Inner 1 / Inner 2 / Inner 3 |')
+      expect(result.content).toContain('| Outer B | Plain |')
+      for (const cell of ['Outer A', 'Inner 1', 'Inner 2', 'Inner 3', 'Outer B', 'Plain']) {
+        expect(result.content.split(cell)).toHaveLength(2)
+      }
+      expect(result.content.match(/\[Table\]/g)).toHaveLength(1)
+      expect(result.metadata?.tableCount).toBe(2)
+    })
+
+    it('separates block elements inside a list item', async () => {
+      const buffer = Buffer.from(
+        `<body><ul><li><div><p>Versions</p><p>Release Information</p></div></li></ul></body>`
+      )
+
+      const result = await parser.parseBuffer(buffer)
+
+      expect(result.content).toContain('• Versions Release Information')
+    })
+
+    it('drops endnote return links but keeps the endnote text', async () => {
+      const buffer = Buffer.from(
+        `<body><p>Body<sup><a href="#endnote-1" id="endnote-ref-1">[1]</a></sup></p>` +
+          `<ol><li id="endnote-1"><p>End text <a href="#endnote-ref-1">↑</a></p></li></ol></body>`
+      )
+
+      const result = await parser.parseBuffer(buffer)
+
+      expect(result.content).toContain('1. End text')
+      expect(result.content).not.toContain('↑')
+    })
+
     it('drops footnote return links but keeps the footnote text', async () => {
       const buffer = Buffer.from(
         `<body><p>Body<sup><a href="#footnote-1" id="footnote-ref-1">[1]</a></sup></p>` +
 
@@ -108,8 +108,8 @@ const NON_CONTENT_SELECTOR = 'script, style, noscript, meta, link, iframe, objec
 /** Block elements inside a table cell; `.text()` would otherwise glue their words together. */
 const CELL_BLOCK_SELECTOR = 'p, div, li, br, h1, h2, h3, h4, h5, h6, tr'
 
-/** mammoth renders a footnote's return link as `<a href="#footnote-ref-N">↑</a>`. */
-const FOOTNOTE_BACKLINK_SELECTOR = 'a[href^="#footnote-ref"]'
+/** mammoth renders a footnote's or endnote's return link as `<a href="#footnote-ref-N">↑</a>`. */
+const FOOTNOTE_BACKLINK_SELECTOR = 'a[href^="#footnote-ref"], a[href^="#endnote-ref"]'
 
 /**
  * Strips the non-content markup and HTML comments from a loaded document so the
@@ -318,7 +318,9 @@ function processListItem(
         return
       }
     }
-    const text = $(child).text().replace(/\s+/g, ' ').trim()
+    const $child = $(child)
+    $child.find(CELL_BLOCK_SELECTOR).after(' ')
+    const text = $child.text().replace(/\s+/g, ' ').trim()
     if (text) ownText.push(text)
   })
 
@@ -335,8 +337,69 @@ function processListItem(
   }
 }
 
+/** A table's own rows: direct `<tr>` children and those under its section elements. */
+function directRows(
+  $: cheerio.CheerioAPI,
+  table: cheerio.Cheerio<AnyNode>
+): cheerio.Cheerio<AnyNode> {
+  return table.children('thead, tbody, tfoot').children('tr').add(table.children('tr'))
+}
+
+/** Nested tables whose nearest enclosing table is the cell's own. */
+function topLevelNestedTables(
+  $: cheerio.CheerioAPI,
+  cell: cheerio.Cheerio<AnyNode>
+): cheerio.Cheerio<AnyNode> {
+  const own = cell.closest('table').get(0)
+  return cell.find('table').filter((_, nested) => $(nested).parents('table').get(0) === own)
+}
+
+/**
+ * The cells of a table in reading order, each rendered with {@link cellText},
+ * for a table nested inside another table's cell.
+ */
+function flattenedTableCells($: cheerio.CheerioAPI, table: cheerio.Cheerio<AnyNode>): string[] {
+  const cells: string[] = []
+  directRows($, table).each((_, row) => {
+    $(row)
+      .children('td, th')
+      .each((_, cell) => {
+        const text = cellText($, $(cell))
+        if (text) cells.push(text)
+      })
+  })
+  return cells
+}
+
+/**
+ * One cell's text on a single line. Block elements inside the cell get a space
+ * so adjacent paragraphs do not glue together — this mutates the live DOM, and
+ * runs before `extractHeadings`/`extractLinks`, so heading or link text inside a
+ * cell gains those spaces too. A nested table is rendered on a clone of the cell
+ * as its cells joined with ` / `, without `[Table]` markers or pipes, so the
+ * outer row stays one line and the inner text appears exactly once.
+ */
+function cellText($: cheerio.CheerioAPI, cell: cheerio.Cheerio<AnyNode>): string {
+  const nested = topLevelNestedTables($, cell)
+  if (nested.length === 0) {
+    cell.find(CELL_BLOCK_SELECTOR).after(' ')
+    return cell.text().replace(/\s+/g, ' ').trim()
+  }
+
+  const clone = cell.clone()
+  topLevelNestedTables($, clone).each((_, table) => {
+    const $table = $(table)
+    $table.replaceWith(` ${flattenedTableCells($, $table).join(' / ')} `)
+  })
+  clone.find(CELL_BLOCK_SELECTOR).after(' ')
+  return clone.text().replace(/\s+/g, ' ').trim()
+}
+
 /**
- * Process table elements to extract structured data
+ * Renders a table as `[Table]`, one `| a | b |` line per row, `[/Table]`. Only
+ * the table's own rows and each row's own cells are visited, so a nested table
+ * contributes to its containing cell (see {@link cellText}) and is never
+ * emitted a second time as rows of its own.
  */
 function processTable(
   $: cheerio.CheerioAPI,
@@ -345,16 +408,14 @@ function processTable(
 ): void {
   contentParts.push('\n[Table]')
 
-  table.find('tr').each((_, row) => {
-    const $row = $(row)
+  directRows($, table).each((_, row) => {
     const cells: string[] = []
 
-    $row.find('td, th').each((_, cell) => {
-      const $cell = $(cell)
-      $cell.find(CELL_BLOCK_SELECTOR).after(' ')
-      const cellText = $cell.text().replace(/\s+/g, ' ').trim()
-      cells.push(cellText || '')
-    })
+    $(row)
+      .children('td, th')
+      .each((_, cell) => {
+        cells.push(cellText($, $(cell)))
+      })
 
     if (cells.length > 0) {
       contentParts.push(`| ${cells.join(' | ')} |`)
@@ -397,7 +458,6 @@ export class HtmlParser implements FileParser {
       const htmlContent = decoded.text
       const $ = cheerio.load(htmlContent)
 
-      // Extract meta information before removing tags
       const title = $('title').text().trim()
       const metaDescription = $('meta[name="description"]').attr('content') || ''
 
 
@@ -1,9 +1,11 @@
 /**
  * @vitest-environment node
  */
-import JSZip from 'jszip'
-import { describe, expect, it } from 'vitest'
+import JSZip, { type JSZipObject } from 'jszip'
+import { afterEach, describe, expect, it, vi } from 'vitest'
+import type { FileParserError } from '@/lib/file-parsers/errors'
 import { extractOpenDocumentText } from '@/lib/file-parsers/odf-text'
+import { MAX_OFFICE_XML_PART_BYTES } from '@/lib/file-parsers/office-text'
 
 const NS =
   'xmlns:office="urn:oasis:names:tc:opendocument:xmlns:office:1.0" xmlns:text="urn:oasis:names:tc:opendocument:xmlns:text:1.0" xmlns:table="urn:oasis:names:tc:opendocument:xmlns:table:1.0" xmlns:draw="urn:oasis:names:tc:opendocument:xmlns:drawing:1.0" xmlns:presentation="urn:oasis:names:tc:opendocument:xmlns:presentation:1.0" xmlns:dc="http://purl.org/dc/elements/1.1/"'
@@ -22,6 +24,10 @@ async function buildOdf(bodyXml: string, extraParts: Record<string, string> = {}
 const text = (body: string) => buildOdf(`<office:text>${body}</office:text>`)
 
 describe('extractOpenDocumentText', () => {
+  afterEach(() => {
+    vi.restoreAllMocks()
+  })
+
   it('drops annotations so the surrounding sentence stays intact', async () => {
     const buffer = await text(
       `<text:p>Aaa <office:annotation><dc:creator>M</dc:creator><text:sender-initials>M</text:sender-initials><text:p>First comment.</text:p></office:annotation>comment ccc.<office:annotation-end office:name="c1"/></text:p>`
@@ -85,6 +91,56 @@ describe('extractOpenDocumentText', () => {
     )
   })
 
+  it('flattens a nested table into its cell without markers', async () => {
+    const cell = (inner: string) => `<table:table-cell>${inner}</table:table-cell>`
+    const p = (t: string) => `<text:p>${t}</text:p>`
+    const inner = `<table:table><table:table-row>${cell(p('In 1'))}${cell(p('In 2'))}</table:table-row><table:table-row>${cell(p('In 3'))}</table:table-row></table:table>`
+    const buffer = await text(
+      `<table:table><table:table-row>${cell(p('Out A'))}${cell(p('Intro') + inner)}</table:table-row></table:table>`
+    )
+
+    const result = await extractOpenDocumentText(buffer)
+
+    expect(result).toBe('[Table]\n| Out A | Intro In 1 / In 2 / In 3 |\n[/Table]')
+  })
+
+  it('rejects an archive without content.xml as invalid_format', async () => {
+    const zip = new JSZip()
+    zip.file('mimetype', 'application/vnd.oasis.opendocument.text', { compression: 'STORE' })
+    zip.file('junk.txt', 'not a document')
+    const buffer = (await zip.generateAsync({ type: 'nodebuffer' })) as Buffer
+
+    await expect(extractOpenDocumentText(buffer)).rejects.toMatchObject<FileParserError>({
+      code: 'invalid_format',
+    })
+  })
+
+  it('returns an empty string for a present but textless body', async () => {
+    expect(await extractOpenDocumentText(await text('<text:p/>'))).toBe('')
+  })
+
+  it('emits an image frame as its alternative text', async () => {
+    const buffer = await text(
+      `<text:p><draw:frame draw:name="Image1"><draw:image xlink:href="Pictures/a.png" xmlns:xlink="http://www.w3.org/1999/xlink"/><svg:title xmlns:svg="urn:oasis:names:tc:opendocument:xmlns:svg-compatible:1.0">Org chart</svg:title></draw:frame></text:p>`
+    )
+
+    expect(await extractOpenDocumentText(buffer)).toBe('[Image: Org chart]')
+  })
+
+  it('rejects a content part above the per-part size cap before parsing it', async () => {
+    const buffer = await text('<text:p>Small</text:p>')
+    const zip = await JSZip.loadAsync(buffer)
+    const entry = zip.file('content.xml') as JSZipObject & {
+      _data: { uncompressedSize: number }
+    }
+    entry._data.uncompressedSize = MAX_OFFICE_XML_PART_BYTES + 1
+    vi.spyOn(JSZip, 'loadAsync').mockResolvedValueOnce(zip)
+
+    await expect(extractOpenDocumentText(buffer)).rejects.toMatchObject<FileParserError>({
+      code: 'complexity_limit',
+    })
+  })
+
   it('includes embedded object content parts after the main document', async () => {
     const buffer = await buildOdf(`<office:text><text:p>Main</text:p></office:text>`, {
       'Object 1/content.xml': `<?xml version="1.0"?><office:document-content ${NS}><office:body><office:text><text:p>Embedded</text:p></office:text></office:body></office:document-content>`,