An Israeli consumer discount skill that maps any store to its best card, club, or voucher deal — covering HTZone, Poalim Wonder, and government price data across 72 stores and 14 categories.

Part of the OpenClaw skill ecosystem.

• "Which card should I use at Victory?" → ranked list of best deals with exact ₪ amounts
• "Where's cheapest for milk?" → supermarket ranking + gov price data reference
• "Any Wolt discounts?" → live voucher prices from verified sources

Hebrew-first. Designed for WhatsApp and Telegram group agents.

discounts.json
├── Structural layer  — which source covers which category, max % off
│                       near-static, update manually (quarterly)
│
└── Deal layer        — exact voucher prices (e.g. Victory ₪300 → ₪259)
                        cached_at timestamp, 30-day TTL
                        refreshed on-demand when stale

Why two layers? Structural data (HTZone covers supermarkets at up to 20%) barely changes. Specific voucher prices drift monthly. Separating them means 80% of queries are answered from near-free static data; only live lookups cost extra.

The agent loads only the relevant category section, not the full KB:

1. Read quick_lookup.store_to_category → identify category (~200 tokens)
2. Read only that category + source metadata (~800–1,200 tokens)

Result: ~75% fewer tokens vs loading the full file.

When cached_at is null or older than 30 days:

1. Answer immediately from structural data (max_pct)
2. If exact price needed → browser fetch → update deals[] + cached_at → save
3. Next request served from cache

Copy to your OpenClaw skills directory:

cp -r clawsavings ~/.openclaw/workspace/skills/clawsavings

The skill is loaded automatically when the agent receives a discount-related query.

A sample of the 24 verified Poalim Wonder deals in the KB:

Check if HTZone or Poalim Wonder benefit structure changed. Update max_pct and last_reviewed.

python scripts/refresh_deals.py --source poalim_wonder

Or manually edit discounts.json: update deals[] array and set cached_at to today.

• Gov XML price integration — product-level price comparison via OpenIsraeliSupermarkets
• HTZone deal automation — browser refresh with saved session
• Max / Isracard — expand coverage to remaining major card programs

clawsavings/
├── SKILL.md              # Agent instructions + decision logic
├── discounts.json        # Knowledge base (structural + deal layers)
├── header.jpg            # Cover image
├── README.md             # This file
└── scripts/
    └── refresh_deals.py  # Deal refresh automation

MIT