Arab Agent Skills

      ###    ####      ###    ####
     #   #   #   #    #   #   #   #
     #####   ####     #####   ####
     #   #   #  #     #   #   #   #
     #   #   #   #    #   #   ####

       ###    ####  #####  #   #  #####
      #   #  #      #      ##  #    #
      #####  #  ##  ####   # # #    #
      #   #  #   #  #      #  ##    #
      #   #   ####  #####  #   #    #

       ####  #   #  #####  #       #        ####
      #      #  #     #    #       #       #
       ###   ###      #    #       #        ###
          #  #  #     #    #       #           #
      ####   #   #  #####  #####   #####  ####

Agent-ready skills for Arab and MENA work. The first public atlas covers regional APIs; the broader ecosystem will include marketing, sales, support, localization, content, operations, and Arab-market context.

Website: ArabSkills.info

This is the public install repo for Arab Agent Skills. It gives agents the regional context they usually miss: local gateways, government rails, webhook gotchas, sandbox notes, compliance flows, Arabic-language nuance, buying norms, and source-backed operating workflows.

Facts are intentionally conservative. If public docs do not prove a claim, vendor files say Unknown from public docs or Needs vendor access.

Origin

Arab Agent Skills started with Arab Payments Skill Atlas, a payments-focused skill built by Mohamed Waleed for one of his projects to help his agents work with regional payment APIs.

Mohamed showed it to Fady Azzouny, and that sparked the bigger question: why not build a maintained skills ecosystem for Arab agents across APIs, marketing, sales, support, localization, content, operations, Arabic NLP, open data, and Arab-market context?

Fady Azzouny and Mohamed Waleed then built Arab Agent Skills from that starting point.

Quick install

npx skills add ArabAgentSkills/Skills

Install a smaller target set when your agent supports filtering:

npx skills add ArabAgentSkills/Skills -a codex -a claude-code -a cursor

What is inside

Path	Purpose
skills/	Agent skills with short SKILL.md routing files, vendor facts, examples, scripts, references, and evals.
data/vendors.public.json	Public vendor registry subset with docs access, confidence, quality score, and source URLs.
data/skills.public.json	Public skill metadata used by indexes and installers.
standards/	Skill style, source quality, release, eval, and security checklists.
docs/	How to use, extend, and review these skills.

Skill Families

• Infrastructure / API skills: payments, logistics, BNPL, communications, e-invoicing/tax, open banking, identity/government, commerce/POS/accounting, HR/payroll, Arabic NLP, and open data.
• Regional Intelligence Skills: country context, Arabic copy and dialect choice, marketing localization, sales playbooks, support tone, product localization, cultural calendars, growth channels, pricing, partnerships, and SEO/AEO/GEO for agent-facing websites.

Regional Intelligence files are AI drafted and source checked where possible. Native review, expert review, and market validation are explicitly marked and not implied.

Country packs included in v0: Egypt, Saudi Arabia, United Arab Emirates, Kuwait, Qatar, Bahrain, Oman, Jordan, Lebanon, Morocco, Tunisia, Algeria, Iraq, Palestine.

Skills

Skill	Family	Category	Priority	Maturity	Coverage	Eval score
Arab E-Invoicing and Tax	Infrastructure	tax	P0	alpha	10 vendors	100
Arab Market Context	Regional Intelligence	market-context	P0	draft	14 countries	100
Arabic Copywriting and Dialects	Regional Intelligence	copywriting-dialects	P0	draft	14 countries	100
MENA BNPL	Infrastructure	bnpl	P0	alpha	7 vendors	100
MENA Communications	Infrastructure	communications	P0	alpha	8 vendors	100
MENA Logistics	Infrastructure	logistics	P0	alpha	15 vendors	100
MENA Marketing Localization	Regional Intelligence	marketing-localization	P0	draft	14 countries	100
MENA Payments	Infrastructure	payments	P0	alpha	15 vendors	100
MENA Product Localization	Regional Intelligence	product-localization	P0	draft	14 countries	100
MENA Sales Playbooks	Regional Intelligence	sales-playbooks	P0	draft	14 countries	100
Arab Identity and Government APIs	Infrastructure	identity-government	P1	alpha	7 vendors	100
MENA Commerce, POS, and Accounting	Infrastructure	commerce-pos-accounting	P1	alpha	8 vendors	100
MENA Cultural Calendar	Regional Intelligence	cultural-calendar	P1	draft	14 countries	100
MENA Customer Support	Regional Intelligence	customer-support	P1	draft	14 countries	100
MENA Growth Channels	Regional Intelligence	growth-channels	P1	draft	14 countries	100
MENA Open Banking	Infrastructure	open-banking	P1	alpha	7 vendors	100
SEO, AEO, and GEO	Regional Intelligence	seo-aeo-geo	P1	draft	16 countries	100
Arab Open Data	Infrastructure	open-data	P2	draft	8 vendors	100
Arabic NLP	Infrastructure	arabic-nlp	P2	alpha	6 vendors	100
MENA HR and Payroll	Infrastructure	hr-payroll	P2	draft	5 vendors	100
MENA Partnerships and Distribution	Regional Intelligence	partnerships-distribution	P2	draft	14 countries	100
MENA Pricing and Packaging	Regional Intelligence	pricing-packaging	P2	draft	14 countries	100

See SKILL_INDEX.md for the full category map and VENDOR_INDEX.md for the vendor list.

Source confidence model

Grade	Meaning
A	Official docs, API reference, OpenAPI/Postman, or SDK evidence is public and useful.
B	Official source exists but is partial, fragmented, gated in places, or needs manual review.
C	Weak public source evidence; useful as a lead but not enough for confident implementation.
unknown	Public evidence is not enough yet.

Source quality score is a review aid, not a certification. It rewards official docs, API references, OpenAPI/Postman assets, official SDKs, and recent evidence.

Safety

• Do not commit API keys, payment credentials, customer data, private docs, or gated vendor material.
• Keep sandbox and production instructions separate.
• Require explicit user approval before live payment, refund, payout, identity, tax, or open-banking actions.
• Do not invent endpoints, credentials, SDK support, pricing, compliance claims, or webhook behavior.
• Do not invent market facts, cultural rules, dialect quality, review status, or local sales/marketing norms.
• Public dialect copy, high-stakes support macros, and strategic regional recommendations require native or expert review unless already approved.
• Prefer official vendor docs and mark uncertainty honestly.

Supported agent surfaces

These skills are written for generic agent-skill runtimes and tested against Codex-style, Claude-style, OpenClaw-style, and generic routing patterns. Each SKILL.md uses short frontmatter plus progressive disclosure so agents load vendor details only when needed.

Contributing

Good contributions add evidence, not guesses. New skills and vendor additions are reviewed before publication; nothing submitted by users or organizations is posted automatically.

1. Add or update official source URLs in the relevant vendor file.
2. Mark docs access and confidence honestly.
3. Keep SKILL.md short; put facts in vendors/*.md.
4. For regional claims, add country, segment, channel, confidence, evidence type, source URLs, and review status.
5. Use standards/source-quality-rubric.md before opening a PR.
6. Add eval prompts when behavior changes.
7. For new skill ideas, open the skill submission issue template and wait for maintainer approval.

Update flow

This repo is regenerated from a reviewed source-of-truth registry. Manual PRs are welcome when they include official sources and keep unknowns explicit.