capture-translate

Capture a web article locally and save a translated markdown version alongside the original.

Updated Apr 23, 2026
One-click install
npx skills add https://github.com/danielrosehill/Claude-Local-Web-Capture-Plugin --skill capture-translate-danielrosehill
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: capture-translate
Source: https://github.com/danielrosehill/Claude-Local-Web-Capture-Plugin/tree/main/skills/capture-translate
Command: npx skills add https://github.com/danielrosehill/Claude-Local-Web-Capture-Plugin --skill capture-translate-danielrosehill

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires libretranslate.

What problem does it solve? Reading geo-restricted or foreign-language articles (especially Hebrew news) requires both capturing the content from your own machine and producing a faithful translation, without sending the text to hosted translation APIs. ## Core Features & Use Cases - Capture-then-translate pipeline: Scrapes the article via the local scrape path first, then translates from the saved markdown so the original is always preserved. - Flexible language pairs: Defaults to Hebrew → English but supports arbitrary source and target languages via user overrides. - Two translator options: Uses Claude inline for single articles, or a locally-running LibreTranslate server for bulk and scripted runs. - Structured output: Saves a stacked markdown file with frontmatter (source URL, languages, translator, timestamps) to <capture-root>/translations/YYYY/MM/. - Use Case: A user shares a Hebrew news article URL and asks for the English version; the skill captures the original markdown and saves a faithful English translation alongside it. ## Quick Start Capture this Hebrew article URL and save an English translation of it.

Frequently Asked Questions about capture-translate

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I translate a Hebrew article to English from a URL?▼

Share the article URL and ask for the English version. The skill first captures the article via the local scrape path, then translates the saved markdown and stores both versions under the translations directory.

What translation tool does this use, Claude or LibreTranslate?▼

Claude translates inline by default for single articles, giving the best quality with no extra dependencies. A locally-running LibreTranslate server is used only for bulk runs, explicit requests, or very long articles.

Can I translate into languages other than English?▼

Yes, the default target is English but any target language can be specified, such as asking to translate to French or Spanish. The source language is also detected from frontmatter rather than assumed to be Hebrew.

Does this send article text to Google Translate or DeepL?▼

No, hosted translation APIs are deliberately not used. Translation happens either inline with Claude or through a LibreTranslate server running on 127.0.0.1, keeping the localhost-only posture intact.

Where are translated articles saved?▼

Translations are saved to <capture-root>/translations/YYYY/MM/ with a filename containing the date, slug, and language pair. The capture root is project-local when inside a git repo, otherwise the global ~/local-web-capture/ directory.