Build your Luna
localisation QA agent

I need to install two skills for Luna: skill-vetter (security checker for skills) and html-text-extract (web page reader). First, use skill-vetter to scan html-text-extract. If it's safe, install both.

I need a tool that checks translated web pages for quality. Here's what I want it to do:

First, research the official OpenClaw documentation on skill creation — check https://docs.openclaw.ai for best practices, metadata format, and how to register skills properly.

Then, think about best practices for translation quality assurance — what do professional reviewers look for? Tone, accuracy, terminology, formatting, UI fit.

Now build a validation tool for Luna. It should work like this:

1. Take source text (English) and translated text (e.g. German) as input — use the html-text-extract skill to pull content from URLs if needed
2. Compare them side by side — flag anything that doesn't look right
3. Check for these issues specifically:
   - Meaning errors — was anything translated incorrectly?
   - Terminology problems — words that should match an approved glossary
   - Cultural adaptation — does it feel natural in the target language, not just translated?
   - Grammar and vocabulary — awkward phrasing, unnatural collocations
   - UI issues — buttons, labels, navigation that look wrong
   - Formatting — numbers, dates, currency in the wrong format
4. Always suggest a fix for every issue you flag — don't just point out problems
5. Output a clean report that I can send to a developer or translator

Store the skill in ~/.openclaw/luna-agent/skills/translation-validation/ — create a SKILL.md with proper metadata (name, description, trigger phrases) and a validation.py script.

Register the skill so Luna can see it — follow the docs for the correct registration flow.

When it's built, test it: use /skill html-extract to pull the first paragraph from https://hypefree-training.pages.dev/ai-writing, then use /skill translation-validation to translate it and validate the output.

I need a readability checker for Luna. When I'm reviewing translated websites, I need to know if the text is too complex for the audience.

First, think about what makes text readable — what metrics do professional editors use? Flesch Reading Ease? Grade level? Sentence length?

Now build a tool that:

Takes a URL or pasted text as input — use html-text-extract if it's a URL
Runs readability metrics on the extracted text and gives me:
How easy is it to read? (Flesch Reading Ease score)
What grade level is it written at? (Flesch-Kincaid)
How many years of education do you need? (Gunning Fog)
A plain-English summary: "Grade 8 — good for general web audiences"
Warns me if the text looks like it's not English (some metrics don't work for other languages)
Warns me if the text is too short for reliable scores (under 100 words)
If the score is poor, suggest what to simplify

Store the skill in ~/.openclaw/luna-agent/skills/readability-check/ — create a SKILL.md and readability_check.py.

Validate readability of this website using the skill: https://hypefree-training.pages.dev/de/ai-writing — compare it against the English original at https://hypefree-training.pages.dev/ai-writing.

Luna, please evaluate these two websites for localisation quality:

German version: https://hypefree-training.pages.dev/de/
English original: https://hypefree-training.pages.dev/

Use all your skills — check translation accuracy, readability, terminology consistency, and UX. Give me a structured report with issues grouped by page section and severity. For each issue, include the original text, the German text, and your suggested fix.

Build a localization file converter for Luna. It should convert between TMX (Translation Memory eXchange) and XLIFF (XML Localisation Interchange File Format) in both directions.

Requirements:
- Auto-detect the input format — if it's TMX, convert to XLIFF. If it's XLIFF, convert to TMX.
- Support TMX 1.4 and XLIFF 1.2
- Handle both single files and entire folders
- Use Python 3 standard library only — zero external dependencies
- Never modify source files
- Output a brief conversion report (files processed, direction, any issues)

Store the skill in ~/.openclaw/workspace/skills/localization-converter/ — create a SKILL.md and converter.py.

Test it with a sample TMX and XLIFF file to verify both directions work.