A Claude skill that adapts content across any language, culture, and platform so it reads native, not translated. Ships a validator.
SaferSkills independently audited vibe-translator (Agent Skill) and scored it 96/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 1 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Most translation moves words from one language to another. This skill moves vibe — the cultural and platform context that makes content actually land.
A perfectly accurate translation of a LinkedIn post into another language will still read as a translated LinkedIn post. It will feel foreign on the destination platform, where the native conventions are completely different. The job here is not translation. It is rebuilding the content as if a native of the target context had written it from scratch.
This works for any language, any culture, any platform, in any direction. The method is universal: the value is in carrying intent across a cultural and platform gap — directness vs. indirectness, self-promotion norms, individual vs. collective framing, humor that doesn't survive a literal hop, formality register, emoji and formatting conventions, and the unspoken "would a local actually post this?" test.
Don't ask "what do these words mean in the other language?" Ask "how would a native of the target context express this same intent?"
This often means changing structure, length, tone, emoji usage, formatting, references, and level of directness — not just vocabulary.
Before adapting, establish (ask only if genuinely unclear):
If the user gives content but no target, infer the most likely one and state your assumption.
Examples, not an exhaustive list. The same method applies to any language or platform not shown here — reason from the cultural dimensions below.
Deliver:
If the target language isn't English, still write the adaptation notes in the language the user is conversing in.
This skill bundles a heuristic checker at scripts/validate.py for the most common targets (xhs, weibo, linkedin, twitter, jp_formal, jp_casual). It catches mechanical "this is secretly a literal translation" smells — missing platform markers, wrong-language output, and structure that mirrors the source 1:1.
python scripts/validate.py --target xhs --source source.txt --adapted adapted.txtFor any language/platform the checker doesn't cover, skip it and rely on the native-fluency gut check — that judgment is always yours.
Input: English LinkedIn post → Xiaohongshu (Chinese)
"Thrilled to announce I've accepted a Product Manager role at ByteDance! Grateful to everyone who supported me. Onwards and upwards! 🚀"
Bad output (translated, not adapted):
"很高兴地宣布我已经接受了字节跳动的产品经理职位！感谢一路上支持我的每一个人。继续向上！🚀"
(Accurate, but reads as a translated LinkedIn post. No one posts like this on XHS.)
Good output (rebuilt for XHS):
啊啊啊我拿到字节的PM offer啦！！！😭🎉
>
还有点不敢相信… 从投简历到面试,真的太多人帮过我了🥹
>
普通人也可以的,真的!如果有姐妹也在准备产品岗,评论区一起聊～
>
\#字节跳动 #产品经理 #offer #求职日记
Adaptation notes:
The same approach works in any direction and any language pair — e.g. a Japanese product update rebuilt with Western directness, a Chinese investing take rebuilt for an English audience, or an English launch post rebuilt for Spanish Instagram.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.