Nadlan Skill

March 29, 2026 · View on GitHub

Nadlan Skill

AI-Powered Israeli Real Estate Investment Analysis

An agent skill that turns any LLM into a real estate data analyst for the Israeli market.
No API keys. No external tools. Just raw HTTP calls to public government APIs.


Govmap Transactions · data.gov.il Open Data · Yad2 & Listing Sites


What Is This?

A self-contained agent skill (knowledge pack) that teaches AI coding assistants how to analyze Israeli real estate deals. Drop it into Claude Code, Cursor, Codex, or any LLM agent that supports skills — and it gains the ability to:

  • Pull actual recorded transactions from the Israeli Tax Authority via Govmap
  • Scrape current asking prices from Yad2, Komo, ad.co.il, and more
  • Cross-reference listings against building-level comparables (not neighborhood averages)
  • Score investment potential using statistical models with outlier detection
  • Generate structured Markdown reports with confidence levels and actionable verdicts

Think of it as giving your AI assistant a real estate analyst's playbook — complete with API specs, algorithms, and methodology.


Workflows

#WorkflowWhat It Does
1Below-Market Deal FinderScrapes listings → compares asking price vs. same-building recorded transactions → flags real discounts
2Property ValuationTakes an address → finds building comps → returns estimated value range with confidence
3Market ComparisonCompares multiple areas side-by-side — price/sqm, appreciation, volume, volatility
4Rental InvestmentPurchase price baseline + rental listings → gross yield with vacancy and risk factors

Data Sources

Govmap — Transaction Records

The core data source. Provides actual recorded sale prices from the Israeli Tax Authority — not estimates, not asking prices. The skill uses four endpoints:

EndpointPurpose
AutocompleteResolve any Hebrew address → ITM coordinates
Deals by RadiusDiscover building polygons near a point
Street DealsRecorded transactions on a specific street/building
Neighborhood DealsWider area transactions as fallback

data.gov.il — Contextual Open Data

Government open data for enrichment — urban renewal zones (Pinui-Binui), construction pipeline, demographics, Mechir LaMishtaken benchmarks, municipal finances, and more.

Listing Sites — Current Asking Prices

Live scraping of for-sale and rental listings:

SiteRole
Yad2Primary source (~80% of listings). HTML + JSON feed + browser fallback
KomoSecondary. Good for cross-referencing
ad.co.ilIncludes AC/renovation details
MadlanWebSearch only (blocked)
HomelessWebSearch only (blocked)

How It Works

Address (Hebrew)


┌──────────────┐     ┌──────────────────┐
│   Govmap     │     │  Listing Sites   │
│  Autocomplete│     │  Yad2 / Komo /   │
│  → Coords    │     │  ad.co.il        │
└──────┬───────┘     └────────┬─────────┘
       │                      │
       ▼                      ▼
┌──────────────┐     ┌──────────────────┐
│  Radius      │     │  Current Asking  │
│  Search →    │     │  Prices + Links  │
│  Polygon IDs │     │                  │
└──────┬───────┘     └────────┬─────────┘
       │                      │
       ▼                      │
┌──────────────┐              │
│  Street &    │              │
│  Building    │◄─────────────┘
│  Transactions│   Compare
└──────┬───────┘


┌──────────────────────────────────┐
│  Outlier Filtering (IQR + %)    │
│  Building-Level Comparison      │
│  Discount Calculation           │
│  Confidence Scoring             │
└──────────────┬───────────────────┘


        Markdown Report
   (with comps, verdicts, links)

Methodology

The skill enforces a strict building-level comparison methodology:

PriorityComparison LevelConfidence
1Same building (address + house number match)HIGH
2Same street (nearby numbers, similar type)MEDIUM
3Neighborhood average (last resort only)LOW

Never claim "below market" from neighborhood averages. The skill is designed to prevent this common mistake that leads to false deal signals.

Outlier Filtering Pipeline

Every baseline calculation passes through a 4-step filter:

  1. Hard bounds — Remove price/sqm outside 1,000–100,000 NIS/sqm
  2. Amount floor — Remove transactions under 100,000 NIS (family transfers, partial sales)
  3. IQR method — Interquartile range with k=1.0 (requires 4+ data points)
  4. Percentage safety net — Remove values >40% deviation from median

Project Structure

israeli-real-estate-analysis/
├── SKILL.md            # Main skill entry point — quick start & rules
├── api-govmap.md       # Govmap API: endpoints, params, response formats
├── api-data-gov-il.md  # data.gov.il CKAN API: datasets, resource IDs
├── api-listings.md     # Yad2, Komo, ad.co.il: URLs, codes, access strategy
├── algorithms.md       # Outlier detection, statistics, investment scoring
└── workflows.md        # Analysis workflows, methodology, report template

Usage

With Cursor / Claude Code / Codex

Place the israeli-real-estate-analysis/ folder in your agent's skill directory and reference it from your skill configuration. The agent will automatically pick it up when users ask about Israeli real estate.

Trigger Phrases

The skill activates on queries mentioning:

Israeli property prices, real estate deals, apartment valuations, investment analysis, nadlan, נדל"ן, דירות, Israeli housing market

Example Prompts

"Find below-market apartments in Holon under 2.5M NIS"

"What's a fair price for a 4-room apartment on Sokolov 38, Holon?"

"Compare investment potential: Bat Yam vs Holon vs Rishon LeZion"

"Analyze rental yield for 3-room apartments in Tel Aviv"

Key Design Decisions

  • No API keys required — All data sources are publicly accessible government APIs
  • No MCP servers needed — Uses raw HTTP requests via whatever fetch tool the agent has
  • Building-level rigor — Prevents misleading "below market" claims from averaged data
  • Graceful degradation — If Yad2 is CAPTCHA-blocked, the skill tells the user instead of producing unreliable analysis
  • Hebrew-native — API calls use Hebrew addresses and place names as required by the data sources
  • Full pagination — Never stops at page 1; best deals often appear on later pages

Supported Cities (Yad2 Codes)

CityHebrew
Tel Avivתל אביב
Jerusalemירושלים
Haifaחיפה
Holonחולון
Ramat Ganרמת גן
Petah Tikvaפתח תקווה
Rishon LeZionראשון לציון
Beer Shevaבאר שבע
Netanyaנתניה
Ashdodאשדוד
Herzliyaהרצליה
Kfar Sabaכפר סבא
Ra'ananaרעננה
Bat Yamבת ים
Rehovotרחובות
Ashkelonאשקלון
Modiinמודיעין
Even Yehudaאבן יהודה

City not listed? The skill will web-search for the correct Yad2 URL codes automatically.


License

MIT