More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Whole-Book Translator, Any Language (PDF, EPUB, DOCX)

    1

    Translates entire books from any language to any language, chunk by chunk, with every sentence, code block, table and figure verified and rebuilt as PDF, EPUB or DOCX.

    Free

    1 useSecurity scanned
    Whole-Book Translator, Any Language (PDF, EPUB, DOCX)

    Whole-Book Translator, Any Language (PDF, EPUB, DOCX)

    Example session with this skill installed

    Translate LinuxSample.pdf (the first 18 pages of "Linux Notes for Professionals") into Spanish. I need a PDF and an EPUB. Formal Spanish, and keep every command and option exactly as it is.

    • Read your context and instructions
    • Compiled the whole-book translator any

    Summary

    Translated Linux Notes for Professionals (sample) into Spanish: 270 segments (about 3,400 words) in 3 chunks, translated by 3 subagents in parallel and checked one by one. Built Linux-Notes-for-Professionals.es.pdf (19 pages) and .epub, then looked at the rendered pages before delivery.

    What was done

    StepResult
    Extraction5 chapters; 69 code blocks copied unchanged; 7 command tables and 45 list items kept as tables and lists
    ExclusionsCover and the printed table of contents left out (a new TOC with page numbers is generated)
    Glossary11 terms fixed before translating (directorio, comando, portapapeles, salida...), 7 names kept as-is
    Translation3 chunks of about 2,000 tokens, one subagent each
    Checkscheck all: 0 errors; every inline command, path and URL accounted for
    ReviewIndependent reviewer read all 272 pairs: 0 critical, 5 minor (term consistency), fixed and re-checked

    Checked before delivery

    CheckResult
    Every source segment translated exactly once✓
    Code blocks identical to the original✓
    Commands, options, paths and URLs unchanged✓
    One Spanish term per glossary entry throughout✓
    TOC page numbers and PDF bookmarks✓

    Next steps

    • Ready for the whole book: the same project continues chunk by chunk; status shows what is left.
    • Need an editable version? Ask for DOCX and it is built from the same translation.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Ask an AI agent to translate a 200-page book and it fails without telling you. The book doesn't fit in its context, so it translates from memory in big pieces. Sentences drop out of long paragraphs, $OFS = " " and URLs disappear, a term like "bucket" is rendered three different ways, tables are flattened into prose, and captions get invented for images nobody looked at. The result reads fluently, so nobody notices.

    This skill gives your agent a tested pipeline that turns the book into numbered segments, translates it in small self-contained tasks, and checks every task against the source before the book is rebuilt.

    What it handles

    • Any input: text-based PDF, EPUB, DOCX, HTML, Markdown and TXT. Chapters come from the outline or headings. Running headers, page numbers and the printed table of contents are left out.
    • Nothing lost: code blocks are copied byte for byte and never pass through the translator. Inline code, URLs, e-mails, link targets and numbers are checked in every segment. Figures are copied as images and never get invented captions.
    • Context-safe chunks: about 2,500 tokens per task, with the rules, style, glossary entries and the previous chunk's last lines included. State lives on disk, so a context reset or a new session continues where it stopped (status).
    • Parallel translation: task next --count 4 hands out four chunks to subagents. One shared glossary and style keep them consistent.
    • Automatic checks: check fails a chunk for missing or misplaced segments, lost code/URLs/terms or untranslated text. It warns on dropped sentences, missing numbers, length outliers and glossary misses.
    • Glossary first: terms finds the recurring vocabulary before translation starts, so a domain word keeps its domain meaning.
    • Real typesetting: PDF is printed through Edge or Chrome, so every script renders correctly, including Chinese, Japanese, Hindi, Thai, Hebrew and Arabic. It includes a TOC with page numbers and bookmarks. EPUB 3, editable DOCX, HTML and Markdown are also available.
    • Review pass: side-by-side review sheets and a fidelity rubric for an independent reviewer agent.

    The page layout is re-typeset, not copied page by page. Text inside images stays as it is. Free and open source (MIT).

    • Scanned PDFs need OCR first; only PDFs with a text layer can be extracted.
    • The book is re-typeset, not reproduced page by page: structure, code, tables, images and links are kept, the original page design is not.
    • Text inside images (screenshots, diagrams) is not translated; images are copied as they are.
    • PDF extraction is heuristic. Very complex layouts (magazines, multi-column pages with sidebars, forms) need options such as --columns 2, or manual exclusions. Check the extraction with show before translating.
    • DRM-protected EPUBs are refused. MOBI/AZW3/DOC must be converted first (Calibre, LibreOffice).
    • Code comments are kept in the source language: code blocks are copied byte for byte.
    • Translation quality is your model's. The skill prevents omissions and drift and makes review systematic, but it does not replace a human editor for publication.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Requires

    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    File Scopes

    */source/**
    */translations/**
    */work/**
    */review/**
    */output/**

    Runs Python scripts locally. It reads the book file you point it at and writes a project folder next to it (source/, translations/, work/, review/, output/). For PDF output it starts Microsoft Edge or Google Chrome headless, only to print a local HTML file. There is no network access and no API keys: your own agent does the translation.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    Unlike standard LLM translations that hallucinate or skip text when context limits are reached, this skill uses a verified pipeline that breaks books into small chunks, cross-references every segment against the source, and preserves technical elements like code blocks and URLs byte-for-byte.

    The skill supports text-based PDF, EPUB, DOCX, HTML, Markdown, and TXT files, and it can export your finished translation into high-quality EPUB 3, editable DOCX, or professionally typeset PDFs with functional tables of contents.

    Yes, the skill includes a 'Parallel Translation' feature that allows you to deploy multiple subagents simultaneously to process different chapters, significantly reducing the time required for long-form books while maintaining consistency through a shared glossary.

    The skill automatically generates a glossary of recurring domain-specific vocabulary before translation begins, ensuring that technical terms remain consistent throughout the entire book even when different chapters are processed separately.

    The package includes the full translation pipeline, the automated fidelity checking system, a side-by-side review sheet generator, and the specialized typesetting engine for professional PDF and EPUB exports.

    While text within the main flow of the book is translated and re-typeset, text contained inside flat images or diagrams remains in its original language to maintain visual integrity and prevent the AI from 'inventing' captions.

    New in Translation & Localization