eval-dataset-employment-prompts-30

Category: Coding Risk: Medium risk ★ 3.9 · Rating 3.9/5 (8) sboghossian/mini-claude-for-legal MIT

Rating is derived from the repo's GitHub stars and shown for reference.

network_access

name: eval-dataset-employment-prompts-30
description: Use when running the employment law benchmark that tests drafting, review, calculation, and advisory capabilities across LB/KSA/UAE/DIFC/UK/EU jurisdictions. Contains 30 prompts spanning contracts, termination, EOSG/EOSA calculations, non-compete analysis, and bilingual Arabic requests.
license: MIT
metadata:
id: eval.dataset.employment-prompts-30
category: eval
priority: P0
intent: [eval, employment, benchmark, dataset, mena]
related: [eval-benchmark-runner, eval-dataset-nda-prompts-30, eval-regression-detector, eval-rubric-legal-soundness, eval-rubric-jurisdiction-awareness]
source: Louis — HAQQ Legal AI (github.com/sboghossian/mini-claude-for-legal)
version: "1.0"

Eval Dataset — Employment Prompts (30)

Scope

30 employment-law prompts designed to benchmark the full range of employment AI capabilities across LB/KSA/UAE/DIFC/UK/EU. Employment law is one of the highest-demand practice areas in MENA (every company needs it), so quality here is directly tied to product-market fit. The EOSG/EOSA calculation prompts are particularly high-value — getting these wrong is an immediate trust failure.

Storage: eval/datasets/employment-prompts-30.jsonl

How to use this pack

Same pipeline as [[eval-dataset-nda-prompts-30]]: load JSONL, run against production endpoint, score via [[eval-llm-as-judge-system-prompt]], aggregate, detect regressions in [[eval-regression-detector]].

Focus: after scoring, check that EOSG calculation prompts scored ≥ 4.0/5 on [[eval-rubric-legal-soundness]]. Calculation errors in this area are high-stakes.

Prompt categories

Category 1 — Drafting (10 prompts)

Contract and letter drafting across jurisdictions:

# Prompt type Jurisdiction Key signals to check
1 Employment contract (full) UAE onshore Probation ≤ 6 months, EOSG clause, Limited vs Unlimited contract
2 Employment contract (full) DIFC DIFC Employment Law compliance, at-will vs cause
3 Employment contract (full) KSA Saudi Nitaqat, Saudization %, 90-day probation
4 Employment contract (full) Lebanon Lebanese Labour Code, NSSF registration
5 Offer letter UAE Non-binding but professional; consistent with planned contract
6 Termination letter (with cause) UAE MOHRE process, 30-day notice (or pay in lieu)
7 Termination letter (redundancy) UK Statutory redundancy, selection criteria, consultation
8 Severance agreement UAE Waiver of EOSG claims; validity under UAE Labour Law
9 Non-compete clause DIFC 12-month max; geographic scope; consideration
10 Fixed-term contract KSA Auto-renewal rules; Saudi Labour Law Article 55

Category 2 — Review (5 prompts)

Paste a contract extract and ask for risk analysis:

# Prompt type Key signals
11 Review UAE employment contract — employer side Identify probation clause trap, missing EOSG formula
12 Review UAE employment contract — employee side Spot overly broad non-compete; missing end-of-service calc
13 Review DIFC employment contract Check for DIFC Employment Law compliance gaps
14 Review non-compete clause — enforceability in KSA KSA courts rarely enforce; flag risk
15 Review restrictive covenants — Lebanon Limited enforceability under Lebanese law; flag

Category 3 — EOSG/EOSA Calculations (5 prompts)

End-of-service gratuity and allowance calculations with edge cases:

# Scenario Key calculation notes
16 UAE: 7 years, terminated without cause, AED 25,000/month Full formula: 21 days × 5 years + 30 days × 2 years; Article 51 UAE Labour Law
17 UAE: 3 years, resigned voluntarily 1/3 of full gratuity for resignation 1–3 years
18 UAE: 6 years, resigned voluntarily 2/3 of full gratuity for resignation 3–5 years
19 DIFC: 5 years, terminated DIFC Employment Law basic wage × years; different formula from onshore
20 KSA: 10 years, terminated 1/2 month per year for first 5 + 1 month per year after; Saudi Labour Law

Category 4 — Advisory (5 prompts)

Practical employment law questions:

# Scenario Key signals
21 "Can I dismiss an employee on sick leave in the UAE?" UAE: protected periods; medical leave rules
22 "What is the process for collective redundancy in the UK?" TULRCA 1992; 45 days for 100+ employees
23 "Does a non-compete in Lebanon prevent my employee from joining a competitor?" Very limited enforceability; case-by-case
24 "Can we put a UAE employee on unpaid leave due to business slowdown?" Requires consent; Article 31 UAE Labour Law
25 "How many annual leave days does a DIFC employee get?" DIFC: 20 days minimum after 1 year

Category 5 — Jurisdiction comparison (3 prompts)

# Prompt Key signals
26 "Compare non-compete enforceability in UAE, KSA, and DIFC" UAE onshore: moderate; KSA: low; DIFC: higher (common law)
27 "How does the probation period differ between UAE onshore and DIFC?" UAE: 6 months max; DIFC: 6 months, different termination rights during probation
28 "What are the EOSG differences between UAE onshore and Abu Dhabi free zone?" May differ by free zone regulations

Category 6 — Bilingual Arabic (2 prompts)

# Prompt Key signals
29 Arabic-language request for UAE employment contract Arabic output, correct legal terminology (عقد العمل، مكافأة نهاية الخدمة)
30 Arabic: "ما هي حقوق موظف في حالة الفصل التعسفي في الإمارات؟" Answer in Arabic, UAE Labour Law reference

Scoring targets

Category Min acceptable score (legal soundness)
Drafting ≥ 4.0 / 5
Review ≥ 3.5 / 5
Calculation ≥ 4.5 / 5 (calculation errors are high-impact)
Advisory ≥ 3.5 / 5
Comparison ≥ 3.5 / 5
Arabic ≥ 3.5 / 5 language quality + ≥ 3.5 legal soundness

Caveats & currency

Employment law in MENA changes frequently. UAE Labour Law (Federal Decree-Law No. 33 of 2021) replaced the previous law; update prompts when further amendments are issued. KSA Labour Law has undergone significant reform since 2021. Review and update this dataset annually or when a major legislative change occurs.

  • [[eval-benchmark-runner]] — runs this dataset in the full eval pipeline
  • [[eval-dataset-nda-prompts-30]] — parallel dataset for NDA tasks (same pipeline)
  • [[eval-rubric-legal-soundness]] — primary scoring rubric
  • [[eval-rubric-jurisdiction-awareness]] — critical for the cross-jurisdiction comparison prompts