- Home
- Skills
- Business & Operations
- Arabic Contact Dedupe: Names, Phones & Emails
Works with the AI Agents you already use
Arabic Contact Dedupe: Names, Phones & Emails
So the CRM keeps two records for one customer and the class list counts one student twice.
$9
Arabic Contact Dedupe: Names, Phones & Emails
Example session with this skill installed
Find the duplicates in this contact list (no file given: run the bundled demo).
- Read your context and instructions
- Compiled the arabic contact dedupe
- Generated the data export
ARABIC CONTACT DEDUPE REPORT v1.0.0
File: bundled sample (samples/contacts-sample.csv)
Encoding: utf-8 Delimiter: , Contacts: 21
Columns used - name: الاسم | phone: رقم الجوال | email: البريد الإلكتروني
SUMMARY
Safe-to-merge groups : 6 (12 rows; contacts left after merging: 15)
Pairs to check by hand: 3
Same name, different contacts (kept apart): 1
WHAT NORMALISING FOUND (counts of rows)
2 names with compound prefixes joined (abd/abu + next word)
1 names with tashkeel (diacritics) removed for matching
1 names carrying invisible characters (bidi marks, zero-width)
8 names with alef/ya/ta-marbuta/hamza-seat variants unified
1 names with Persian/Urdu keheh or yeh mapped to Arabic letters
1 names with tatweel (kashida) removed
1 phone cells written with Arabic-Indic digits
SAFE TO MERGE - a shared phone or email AND compatible names for every row in the group
G1
row 2 محمد عبد الله الشاذلي | 01001234567 | m.shazly@example.com
row 3 مُحمّد عبدالله الشاذلى | 010 0123 4567 | -
why rows 2 + 3: same phone, and the same name once spelling variants are unified
G2
row 4 احمد السيد | ٠١٢٢٣٣٤٤٥٥٦ | ahmed.sayed@example.com
row 5 أحمد السـيد | 00201223344556 | AHMED.SAYED@example.com
why rows 4 + 5: same email, and the same name once spelling variants are unified
G4
row 10 خالد بن سعيد العمري | 0555000111 | -
row 11 خالد سعيد العمري | 0555000111 | -
why rows 10 + 11: same phone, and the same name once bin/ibn/al are ignored
G5
row 18 نورا حسن محمود | 0123456780 | -
row 19 نورا حسن | 0123456780 | -
why rows 18 + 19: same phone, and one name is a shorter form of the other
(G3 and G6 omitted here: abdel rahman spacing, Persian ک)
CHECK BY HAND - the evidence is not enough to merge
rows 6 + 7: فاطمة الزهراء علي / فاطمه الزهراء على
why: phone matches on the last 9 digits only (the full numbers differ, often a missing country code - pass --country to confirm), and the same name once spelling variants are unified
rows 12 + 13: يوسف إبراهيم / سارة إبراهيم
why: same phone but different names - often a family or office number; do not merge without checking
rows 16 + 17: عمر فاروق [has invisible characters] / عمر فاروق
why: the same name once spelling variants are unified, and one side has no phone or email to confirm - check by hand
KEPT APART - same name, different phone or email (probably different people)
rows 14 + 15: محمد أحمد / محمد احمد
contacts-sample.csv
CSV · data export
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
Ordinary dedupe tools compare text. In Arabic the same person is typed five ways that look alike and compare different: مُحمّد / محمد, أحمد / احمد, فاطمة / فاطمه, علي / على, عبد الله / عبدالله, a hidden right-to-left mark left by a copy-paste, a phone written ٠١٠٠١٢٣٤٥٦٧. So the CRM keeps two records for one customer and the class list counts one student twice.
The opposite mistake is worse. Arabic has many very common full names, and a tool that merges on the name alone folds two different people called محمد أحمد into one record.
What you get
A report that sorts every candidate pair into three lists, each with a one-line reason:
- Safe to merge: a shared phone or email, and names that agree for every row in the group.
- Check by hand: a shared family or office number with different names, a phone that matches only without its country code, or a name match with nothing to confirm it.
- Kept apart: the same name with different contacts, which usually means two people.
Optionally you also get an annotated copy of your CSV. It keeps every original row and cell and adds dup_group, dup_check_with_rows and the matching keys, ready to filter in Excel or Google Sheets.
How it works
A Python script, standard library only, that runs locally with no network:
- Name matching key. It removes tashkeel and tatweel, unifies the alef and hamza-seat forms, ta marbuta / ha and alef maqsura / ya, maps Persian ک and ی to Arabic letters, strips invisible bidi and zero-width characters, joins
عبد/abd elcompounds, and ignores bin, ibn and al where it compares names. - Phones. It reads Arabic-Indic digits,
+,00,(0), extensions and several numbers in one cell. Pass--countryfor local numbers (21 countries: Gulf, Levant, Egypt, Maghreb, Iraq, Yemen, Sudan, Turkey, UK, US). - Grouping. Each group is built so that no two people with different names can end up in it, even when three rows share one family phone.
- Columns. English and Arabic headers are recognised, and Google Contacts and Outlook exports work as exported.
Running it with no file gives a full, labelled demo on a bundled 21-row sample.
Limits
- It does not transliterate (
Mohamedis not linked toمحمد) and does not know nicknames. - It never merges or deletes anything; you merge in your own tool.
- It cannot tell apart two people who share both a name and a phone.
- Excel files must be saved as CSV first.
Needs Python 3.8+ and a CSV (save Excel files as CSV first). No transliteration between Arabic and Latin script, no nickname or kunya mapping, no fuzzy edit distance on names. Cannot tell apart two people who share a name and a phone. Phone country rules cover mobile number lengths for 21 countries; other numbers are compared as typed. Grouping of a three-way tie can depend on row order (no group ever holds two different names). A Western-European (cp1252) file is read as Windows-1256 with a warning; re-save it as UTF-8.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 4 days ago
Requires
Permissions
File Scopes
No network calls. Contact data never leaves your machine.