skill

paper-translate-agent-skill

Portable Agent Skill for active-chat-model PDF translation with resumable batches and layout-preserving rendering

AIKONG2024/paper-translate-agent-skillcommunitylistedunreviewed

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/AIKONG2024/paper-translate-agent-skill itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 21, 2026

paper-translate-agent-skill

English | 한국어

Translate local, text-based PDFs in your Agent chat while preserving the original page layout. The active chat model performs the translation; the bundled scripts prepare the PDF, save resumable batches, rebuild the document, and validate the result.

Set it up in one chat message

Paste this into Codex App or another Agent that can install GitHub skills:

Install and configure the pdf-translate Skill from this GitHub repository:
https://github.com/AIKONG2024/paper-translate-agent-skill/tree/main/pdf-translate

When the Agent finishes setup, start a new chat if requested. Then attach a PDF and ask:

Translate this PDF into Korean and return a PDF that preserves the original layout.

That is the complete user workflow. You do not need to configure a model API key or copy an OAuth token. The Agent already running in the chat translates the document.

Before and after

Last page of YOLOv3: An Incremental Improvement (arXiv:1804.02767v1), translated to Korean by the active chat model:

The example was produced through the installed Skill workflow and passed page geometry, nonblank-content, and preview validation.

What the Skill does

Extracts ordered translation units from the requested PDF pages.
Lets the active chat model translate bounded JSON batches.
Saves accepted batches to ordinary files so interrupted work can resume.
Rebuilds translated-only or bilingual PDFs through the pinned local layout engine.
Publishes only after structural and visual-preview checks pass.

The scripts do not call a provider API, invoke another model process, reuse host OAuth credentials, run an MCP server, or store jobs in a database.

Compatibility

Codex App has been demonstrated end to end. Codex CLI, Claude Desktop, OpenCode, and other Agents can use the same Skill when they support local GitHub skills, local PDF access, UTF-8 file read/write, Python 3.11 or 3.12, shell commands, and first-run dependency installation. Language quality depends on the active model selected in that host.

For manual installation, copy the complete pdf-translate/ directory to the host's skills directory. Install the subdirectory containing SKILL.md, not the repository root.

Repository contents

pdf-translate/: the installable Skill.
tests/: regression and security tests for maintainers.
typings/: narrow static-analysis stubs used by repository checks.

User PDFs, translation jobs, rendered outputs, runtimes, credentials, and logs are intentionally excluded. See LICENSE and THIRD_PARTY_NOTICES.md for licensing and upstream attribution.