Case study · Bilingual publishing
30+ airline promo pages, EN + AR, from a single bulk run.
A Claude-driven publishing skill that replaced a three-tool manual juggle with one unattended run straight into Prismic’s slice schema — both languages at once.
01 — The problem: a three-tool juggle
Publishing airline promo-code pages in Arabic and English meant moving between three separate tools, doing per-page manual steps in each, and constantly context-switching between languages. It was slow, repetitive, and exactly the kind of work that drifts in quality when a person is doing it forty times in a row.
02 — The publishing skill
I built a Claude-driven publishing skill that takes a structured brief, drafts both language variants with Wego’s tone-of-voice applied, and writes directly into Prismic’s slice schema — not a flat blob, but the real structured content model the site renders from. One trigger produces both the English and Arabic variants of a page, correctly structured, ready to publish.
03 — Why a reusable skill, not a one-off script
I built this as a skill — versioned, reusable, parameterised — before Prismic shipped their own MCP integration. Being early meant I owned the mapping between a brief and Prismic’s schema directly, which is exactly what made bulk runs safe: the same tested path every time, not a fresh improvisation per page.
04 — The bulk run
The payoff is throughput. One bulk run publishes 30+ airline promo pages across EN + AR with zero per-page manual steps, fully unattended. The content team stopped being a publishing bottleneck and went back to deciding what to publish rather than operating three tools to ship it.
The honest footnote: Prismic later shipped their own MCP, which would shorten parts of this build today. That’s the nature of being early — the win was shipping a working bilingual bulk pipeline months before the off-the-shelf path existed.
Takeaways
- Write to the real schema, not a flat blob — structured output is what makes bulk publishing safe and reusable.
- Package repeatable work as a versioned skill, not a throwaway script; the second run is where it pays off.
- Bilingual isn’t a translation step bolted on at the end — it’s cheaper to generate both variants from one brief than to localise after the fact.