Arabic AI Skills
Production-focused skills for natural Arabic UX writing, RTL interfaces, product localization, and Arabic copy review across Codex, Claude Code, Cursor, and other AI agents.
مهارات مفتوحة المصدر لكتابة واجهات عربية طبيعية، ومراجعة تجربة RTL، وتوطين نصوص المنتجات.
📌 Versioning Alignment
Framework Spec: v1.2.0-STABLE | Skill Package: v1.2.3 | API Compatibility: 1.x
❓ What is this?
Arabic Intelligence is a guided 11-stage workflow (stages 0–10) for generating natural, culturally localized Arabic UX microcopy, product copy, and interface messaging.
💡 Why does it exist?
Standard LLMs can struggle with Arabic product copy—defaulting to literal English translations, robotic administrative clichés ("في إطار حرصنا المستمر", "يرجى التكرم بالعلم"), and high cognitive load. Arabic Intelligence reduces these patterns through structured context, domain guidance, anti-examples, and an 18-metric review rubric.
👤 Who is it for?
- AI Agent Engineers & Prompt Designers: Adding Arabic product copywriting capabilities to agents.
- Product Managers & UX Writers: Standardizing localized Arabic UX copy across platforms.
- GovTech, FinTech & E-Commerce Teams: Enforcing authentic regional phrasing (Saudi & Gulf).
🎬 See the Difference
▶ Watch the full 54-second HD demo — three before/after transformations covering GovTech empty states, payment errors, and location-permission prompts.
⚡ Quick Example
❌ Raw AI Output (Robotic Cliché Slop)
"في إطار حرصنا المستمر على تقديم أفضل الخدمات، يرجى التكرم بالعلم بأنه لا يوجد لديكم أي مخالفات مرورية حالياً."
✅ Arabic Intelligence Output
العنوان: "لا توجد مخالفات مسجّلة عليك" الوصف: "لا توجد أي مخالفات مرورية مسجلة بحقك حالياً. نتمنى لك قيادة آمنة."
🚀 Quick Install
Choose the focused writing framework or the combined UI/RTL skill:
npx skills add theonlym7md/arabic-ai-skills --skill arabic-intelligence
npx skills add theonlym7md/arabic-ai-skills --skill arabic-uiux-master
Browse the published collection on skills.sh or open either listing: arabic-intelligence · arabic-uiux-master.
| Skill | Best for |
|---|---|
arabic-intelligence |
Arabic product copy, localization, terminology, and copy audits |
arabic-uiux-master |
Arabic interface copy combined with RTL, typography, and accessibility guidance |
Optional local Node.js SDK
The repository includes an SDK for contributors and local integrations. It requires Node.js 22
or 24 LTS and an OPENAI_API_KEY environment variable.
git clone https://github.com/theonlym7md/arabic-ai-skills.git
cd arabic-ai-skills
npm ci
npm run build
const { generateArabicCopy } = require("./dist/index.js");
async function main() {
const result = await generateArabicCopy({
projectNiche: "GovTech",
targetAudience: "Saudi residents",
toneOfVoice: "Official and clear",
textType: "Empty state",
context: "Traffic violation page with no recorded violations",
});
console.log(result.options);
}
main().catch(console.error);
The SDK reads OPENAI_API_KEY from the environment by default. Set OPENAI_MODEL or pass model to choose a model explicitly.
Responses are not stored by default by this SDK; pass store: true only when later retrieval is required. For reproducible production behavior, set model to a dated model snapshot supported by your account.
The same example is available as examples/sdk.cjs.
🧠 Architecture Topology
graph TD
S0[Stage 0: Planning] --> S1[Stage 1: Diagnose]
S1 --> S2[Stage 2: Intent]
S2 --> S3[Stage 3: Audience & Domain]
S3 --> S4[Stage 4: Risk Assessment]
S4 --> S5[Stage 5: Knowledge Selection]
S5 <--> S6[Stage 6: Reasoning & Draft]
S6 --> S7[Stage 7: Rule Weights]
S7 --> S8[Stage 8: 18-Metric Evaluation]
S8 -- Metrics Pass --> S10[Stage 10: Final Output]
S8 -- Metrics Fail --> S9[Stage 9: Dynamic Repair]
S9 -- Re-draft Loop --> S6
⚖️ Capability Matrix
| Category | Status | Capabilities & Directives |
|---|---|---|
| GovTech & Official | Production / Stable |
Saudi & Gulf GovTech tone, empty states, official notices |
| FinTech & Payments | Production / Stable |
Payment friction removal, checkout retry, trust anchors |
| SaaS & Landing Pages | Production / Stable |
Hero value propositions, CTAs, feature cards |
| E-Commerce | Production / Stable |
Cart recovery, transactional alerts, product copy |
| Media & Content | Production / Stable |
Podcasts, White Dialect Arabic, youth-focused copy |
| Healthcare | Experimental |
Patient onboarding, appointment booking |
| Legal Contracts | Unsupported |
Not supported. Requires specialized legal counsel. |
| Medical Diagnostics | Unsupported |
Not supported. Requires licensed medical evaluation. |
| Marketing Spam | Unsupported |
Forbidden by framework ethical rules. |
📂 Protocol Specifications & Developer Tools
- 📜 Protocol Specification: Read SPEC.md for formal stage contracts, error codes, & runtime SDK specs.
- 📑 Compatibility Spec: Read docs/COMPATIBILITY.md.
- 📑 CI Pipeline Specification: Read docs/CI_PIPELINE.md.
- 📑 Architecture Decision Records: Read ADR-001, ADR-002, and ADR-003.
- 📊 Benchmark Methodology: Read benchmarks/README.md for rubric evaluation details.
- 🛠️ Universal System Validator: Run
node scripts/validate_all.js. - 📦 Pristine Clean Packaging: Run
node scripts/package_zip.jsto build a clean distribution zip (excluding.git&node_modules). - 🤝 Contributing: Check our CONTRIBUTING.md guide before submitting PRs.
- 🔒 Security: Report vulnerabilities privately using SECURITY.md.
- 🧭 Roadmap: See ROADMAP.md for focused future work.
- 📝 Releases: Review user-visible changes in CHANGELOG.md.
Stable 1.x framework specification. Changes are tested on Node.js 22 and 24 LTS.

No comments yet
Be the first to share your take.