majdghithan/agent-skills / skills/lahja
lahja
VerifiedArabic dialects (لهجة) writing skill - write or rewrite Arabic in a specific spoken dialect with correct dialect GRAMMAR, not just sample words. Covers negation, future and aspect markers, demonstratives, question words, spelling habits, MSA leaks to remove, and how English tech terms sit inside dialect text. Covers Levantine (Palestinian-anchored), Egyptian, Gulf, Iraqi, Maghrebi (Moroccan, Algerian, Tunisian), Libyan, Sudanese, and Yemeni. Use when the user asks to write "بالعامية", "باللهجة الفلسطينية/الشامية/المصرية/الخليجية/العراقية/المغربية/الليبية/السودانية/اليمنية", "in Levantine/Palestinian/Egyptian/Gulf/Khaleeji/Iraqi/Moroccan/Libyan/Sudanese/Yemeni Arabic or Darija", "make this sound less formal/less translated", "convert فصحى to عامية", or writes to you in a dialect and wants dialect output (social posts, chat/WhatsApp copy, casual product copy, scripts). Built on the fasih method by Maher El Gamil. Works standalone; if the separate fasih skill is also available, invoke it for editorial quality while lahja owns the dialect forms. Do NOT use for formal MSA writing, legal/government/medical text, literal translation, or when no dialect is wanted.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 12.2 KB |
| references/egyptian.md | markdown | 11.1 KB |
| references/gulf.md | markdown | 12.0 KB |
| references/iraqi.md | markdown | 12.5 KB |
| references/levantine.md | markdown | 16.0 KB |
| references/libyan.md | markdown | 13.2 KB |
| references/maghrebi.md | markdown | 15.6 KB |
| references/sudanese.md | markdown | 11.5 KB |
| references/yemeni.md | markdown | 12.8 KB |
| scripts/check_output.py | script | 3.7 KB |
| scripts/check_skill.py | script | 4.1 KB |
references/libyan.md
View on GitHub# Libyan Arabic (ليبي) - writing guide
**Coverage:** Tripoli / western Libyan, the media standard, with Benghazi / eastern notes. Every form below is cited to the sources at the end; anything the sources didn't settle is listed in the last section and must not be generated.
**Evidence tags:** unmarked = published source. `[corpus]` = counted in a 4,061-segment corpus of spontaneous Libyan speech in Arabic script, so it is attested usage rather than a grammarian's claim. Eastern forms are one notch less certain throughout, because the eastern reference grammars are offline.
## 0. Libyan is not Tunisian and not Egyptian. Eight divergences
This is the whole reason the guide exists. An agent that reaches for a neighbour gets these wrong:
| Point | Neighbour | Libyan |
|---|---|---|
| Habitual "I say" | Egyptian بقول, Moroccan كنقول | **بare imperfect: نقول.** Libyan has **no habitual preverb at all**. [Sellami] |
| بـ + verb | Egyptian habitual | **intention future**: بنقول = "I'm going to say", never "I say". [Pereira] [Algryani] |
| باش | Tunisian and Algerian future ("will") | **purpose only**: باش نقول لهم = "so that I tell them". [corpus] |
| "what" | Egyptian إيه, Moroccan آش | **شن** (also شنو) [corpus] |
| "very / a lot" | Tunisian برشا, Algerian ياسر, Egyptian أوي | **هلبا** (west), **واجد** (east) [ArXivLid] |
| "now" | Moroccan دابا, Egyptian دلوقتي, Levantine هلق | **توّا** [corpus] |
| Verbal ـش | obligatory in Egyptian and Moroccan | **optional in eastern Libyan**: ما كتب (ش) [Owens via Algryani] |
| "of" | Moroccan ديال, Egyptian بتاع | **نتاع** / **متاع** [corpus] |
## 1. Negation
- **Verbal negation = ما...ش** around the verb, past and present: علي ما كتبش الرسالة، ما نعرفش، ما نقدرش، ما عندهاش. [Algryani] [corpus]
- **Eastern Libyan may drop the ـش**: ما كتب. Don't do this in a Tripoli text. [Algryani citing Owens]
- **Nominal negation = مش**, standing between subject and predicate: عمر مش هنا، مش علي اللي جا. Write مش, not موش (zero corpus hits). [Algryani] [corpus]
- **Negated future uses مش, not ما...ش**: علي مش حيسافر بكرة. This is the most-missed rule in the dialect. [Algryani]
- **Pronoun copula negation = ما + pronoun + ش**: ماهوش، ماهيش، ماهمش، مانيش. نادية ماهيش هنا. [corpus]
- **Existential**: positive في / فيه; negative **ما فيش** (also written مافيش). Past: ما كانش، ما كانتش، ما كانوش. [corpus]
- **Prohibition = ما + 2nd-person imperfect + ش** (not the imperative stem): ما تمشيش، ما توقفش، ما تتكلمش. [corpus]
- **نبي "want" negates the same way**: ما نبيش، ما نبوش، مابيش. [corpus]
- **Contrast also takes مش**: عمر مشى للسينما، مش للجامعة. [Algryani]
## 2. Future: four markers, not one
Libyan has a layered future. Pick by intent. [Pereira] [Algryani] [corpus]
| Marker | Meaning | Example |
|---|---|---|
| **حـ / حاـ** | certainty, the neutral future; commoner among younger educated speakers | حيكون، حتوصل، حنشوفو |
| **بـ** | intention | بنقول، بندير، بنشوفك غدوة |
| **توـ / توّـ** | commitment or threat, **1st person with a human subject only** | تونورّيك، توندوّرلك |
| **راح + imperfect** | motion-participle future, less frequent | راح نسافر |
- The bare imperfect also carries future in subordinate and hypothetical clauses. [Pereira]
- **باش is not a future marker in Libyan.** It means "so that". [corpus]
- **Not sourced:** ني as a future marker. It appears in no source and nowhere in the corpus.
## 3. Aspect
- **Bare imperfect = habitual or generic**: ياخدوا السحلب ويطحنوه في طاحونة. Libyan has **no** Egyptian بـ- and **no** Moroccan كـ- habitual. [Sellami]
- **The progressive must be marked.** Two ways: [Sellami] [Pereira]
- **قاعد / گاعد + imperfect**, agreeing (قاعد / قاعدة / قاعدين): الشغل اللي قاعد يحدث، أنا گاعدة نطيّب في مكرونة. Here قاعد does **not** mean "sitting".
- **في between verb and object**: نشرب في شاهي = "I'm drinking tea". This في is an aspect marker, not the preposition.
- **راهو / راهي / راني / راهم** is a real presentative-assertive particle in Libyan speech: هذا سوق راهو. Treat it as emphasis, not as the progressive marker. `[corpus]`
- **Active participle = imminent action**: ماشية "I'm off". [Pereira]
- **Bare imperfect after modals**, no linker: نبي نجي، يقدر يكتب رسالة، ضروري علي يجي. [Algryani]
- **Past habitual and past progressive are both كان + imperfect**: علي كان يدخّن. [Algryani]
## 4. Demonstratives and question words
- **Near**: هذا / هاذا (m), هذي / هاذي (f), هذو / هذوما / هذوم (pl). **Far**: هذاك / هذيك، plural هذين / هذينا. Emphatic: هذايا، هذيا. [Pereira] [corpus]
- Write **هذي**, not the MSA هذه, in dialect text. [corpus]
- **Question words**: [corpus] [Algryani] [ArXivLid]
| Meaning | Libyan |
|---|---|
| what | **شن**, also شنو (m) / شني (f, western) |
| who | من / منو / مني |
| where | وين، لوين ("to where")، منين ("from where") |
| why | علاش (west) / ليش (east and centre) |
| how | كيف (**never كيفاش**) |
| when | امتى (**never متين**) |
| how much | قداش / گداش, also كم |
| which | أي / أيّا + noun |
- **Relative = اللي**, overwhelmingly (656 corpus hits vs 5 for الذي). [corpus]
## 5. Pronouns: Libyan is Maghrebi here
- **The n- pattern applies**: **نكتب = "I write"**, **نكتبوا = "we write"**. An agent that writes أكتب or Egyptian بكتب for "I write" is instantly not Libyan. This is the single most important form in the guide. [LibyaWiki] [Algryani] [corpus]
- **Independent pronouns**: اني (the characteristic Libyan "I", commoner than أنا), انت، انتي، هو، هي، **احنا** (Tripoli also احني; eastern and central نحنا)، انتوا، هما. Avoid انتم, which is an MSA leak. [Algryani] [corpus]
- **Possessive particle: نتاع** (commonest) or **متاع**, with a feminine-agreeing متاعت- before a feminine possessor: الإنجازات نتاعنا، الشنطة متاعتها. حق exists but is marginal, and ديال / تاع are not Libyan. [corpus] [ArWikiLibya]
## 6. Spelling habits
- **ق is pronounced [g] but written ق** in ordinary text. گ is also genuine Libyan practice when someone is writing pronunciation deliberately (گالوا، حايگون), but the corpus is 3,722 ق against 687 گ. Default to ق and stay consistent. [LibyaWiki] [corpus]
- **ج is generally [ʒ]**; the western sibilant-harmony spellings (زوز for جوز) are speech, not standard writing. [ArXivLid]
- **Interdentals split west/east**: Tripoli merges ث ذ ظ into stops, Benghazi keeps them. **Spelling does not follow the merger**: write هذا، هذي، كذاب with their etymological letters even for Tripoli. [LibyaWiki] [EastLibyaWiki] [corpus]
- **Final ة** stays ة. Converting it to ه is NLP preprocessing, not Libyan writing. [ArXivLid]
- **Write Arabic script.** The corpus has zero Latin-script tokens. [corpus]
- **Not sourced:** any Libya-specific Arabizi convention, including how [g] gets romanized.
## 7. MSA leaks - kill on sight
| Leak (MSA) | Libyan | Source |
|---|---|---|
| سوف / سـ | حـ (certain) / بـ (intention) / توـ (1sg commitment) | [Pereira] |
| لم / لن | ما...ش | [Algryani] |
| ليس | مش / ماهوش | [Algryani] [corpus] |
| الذي / التي | اللي | [corpus] |
| هذه | هذي | [corpus] |
| ماذا / ما | شن / شنو | [corpus] [ArXivLid] |
| لماذا | علاش (west) / ليش (east) | [ArXivLid] |
| أين | وين | [corpus] |
| كيف | كيف (unchanged) | [Algryani] |
| متى | امتى | [Algryani] |
| الآن | **توّا** / توا / تو | [corpus] |
| أيضاً | زادة / حتى / برضو | [Algryani] |
| جداً، كثير | **هلبا / هلبة** (west), **واجد** (east). Note جداً is alive in Libyan too (92 corpus hits), so it is not always a leak | [ArXivLid] [corpus] |
| يريد | نبي / يبي / تبي | [Algryani] [corpus] |
| يستطيع | نقدر / يقدر | [Algryani] |
| فقط | بس، غير | [corpus] |
| لا يوجد | ما فيش | [corpus] |
| يفعل | **دار** (avoid سوّى and يفعل) | [Algryani] |
| جيد | باهي / باه, quintessentially Libyan | [ArXivLid] |
| هكذا | هكي | [corpus] |
| ذهب | مشى (west) / عدّى (east); imperative برّا (west) | [ArXivLid] |
**West vs east lexicon** worth keeping straight: حوايج / دبش (clothes)، حقّيته / شبحته (I looked at it)، بورزة / شكارة (bag)، صنّة / بنّة (smell)، ماكلة / وكال (food)، صقع / سميطري (cold). [ArXivLid]
## 8. Foreign tech terms: Italian legacy, English present
- **Italian is the historical layer, English is the live one.** Libya is the only Maghreb country whose European contact language was Italian rather than French. Italian loans are about 93% nouns and mostly material vocabulary: كمبيو (gearbox)، فوركيتة (fork)، مرشابيدي (sidewalk). Few Libyans under about 60 speak Italian now. [Benkato]
- **Loan verbs exist but the class is closed**: فريز "to break down", فران "to brake". Don't coin new ones. [Benkato]
- **Turkish layer**, mostly domestic nouns plus دغري "straight ahead" and بلكي "maybe". [Benkato]
- **English tech terms are written in ARABIC script** in real Libyan text: اوكي، ميديا، تيم، سيستم، بوست، شير، بروجكت، كلاس، ايميل، كمبيوتر، انترنت. The corpus has **zero** Latin-script tokens. [corpus]
- **This is the opposite of the Levantine and Gulf default.** In a Libyan text, transliterate the tech term instead of leaving it in Latin, unless the user's own text uses Latin script.
## 9. Worked rewrites (MSA -> Tripoli)
Grammar comes from the sections above; ordinary content words are not individually sourced.
**1. Habitual vs intention** (the بـ trap)
- MSA: أكتب التقارير كل أسبوع، وسوف أرسل التقرير اليوم.
- Libyan: نكتب التقارير كل أسبوع، وبنبعت التقرير اليوم.
- Bare نكتب for the habit, بـ for the intention. Egyptian بكتب would be wrong for the first.
**2. Negated future** (the مش rule)
- MSA: لن نطلق التحديث اليوم، لا يوجد وقت.
- Libyan: مش حنطلقو التحديث اليوم، ما فيش وقت.
- Negated future takes مش, not ما...ش.
**3. Progressive + what** (obligatory progressive marking)
- MSA: ماذا تفعل الآن؟
- Libyan: شن قاعد تدير توّا؟
- Bare تدير would read as habitual, so قاعد is required.
**4. Want + can't** (نبي, n- prefix, tech term in Arabic script)
- MSA: أريد أن أفتح البرنامج لكنني لا أستطيع.
- Libyan: نبي نفتح البرنامج، بس ما نقدرش.
## 10. Deliberately NOT included
Not attested, or actively contradicted, in the sources checked: **متين** for "when" (the corpus hit is the adjective "solid"; use امتى); **ني** as a future marker; **كيفاش** and **وقتاش** (Tunisian, Algerian, Moroccan); **شكون** (Maghrebi, use من / منو); **هاذوكم**; **موش** (write مش); **فين** (Egyptianism); بـ as a habitual marker; باش as a plain future; newly coined Italian or English loan verbs; Libya-specific Arabizi rules.
## Feedback
**Found something wrong or missing?** Arabic dialects vary by city, generation, and family, and this guide will not match every speaker. Tell Majd: majd.ghithan20@gmail.com, or open an issue or PR at https://github.com/majdghithan/agent-skills.
## Sources
- [LibyaWiki] https://en.wikipedia.org/wiki/Libyan_Arabic · [EastLibyaWiki] https://en.wikipedia.org/wiki/Eastern_Libyan_Arabic · [ArWikiLibya] https://ar.wikipedia.org/wiki/لهجة_ليبية
- [Pereira] Pereira, Les futurs en arabe de Tripoli (Libye) - https://books.openedition.org/iremam/4869?lang=en
- [Algryani] Algryani 2012, The Syntax of Ellipsis in Libyan Arabic (Newcastle PhD; ch. 2 is a western Libyan grammar sketch) - https://theses.ncl.ac.uk/jspui/bitstream/10443/1622/1/Algryani%2012.pdf
- [Sellami] Sellami, The Typology of Progressive Constructions in Arabic Dialects - https://cpb-us-w2.wpmucdn.com/voices.uchicago.edu/dist/2/2933/files/2021/06/Sellami_Typology_Final_word.pdf
- [Benkato] Benkato, Maghrebi Arabic, in Arabic and Contact-Induced Change (Language Science Press) - https://langsci-press.org/catalog/view/235/1811/1837-1
- [ArXivLid] Computational Linguistics Meets Libyan Dialect - https://arxiv.org/pdf/2512.04257
- [MADAR] Bouamor et al. 2018, MADAR corpus (includes Tripoli and Benghazi) - https://aclanthology.org/L18-1535.pdf
- [corpus] lahgtna Libyan speech transcripts, Arabic script, counted for this guide - https://huggingface.co/datasets/oddadmix/lahgtna-libyan-tts
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |