hunt-pricing-anomaly

name	hunt-pricing-anomaly
description	On a multi-variant AliExpress listing, find pricing anomalies — where price doesn't scale with volume or quantity, so a bigger size or a bulk tier is disproportionately cheap. Computes ₪-per-litre per variant and reads quantity tiers.

Hunt Pricing Anomaly

Many AliExpress listings reuse one price across several sizes/variants, or hide a steep bulk tier. That means the biggest variant, or a quantity break, can be disproportionately cheap — you get far more product per shekel without paying proportionally more. This skill finds those anomalies on a listing.

When to use

• A listing has a size / capacity dropdown (multiple SKUs under one item).
• You're buying a commodity where more volume is strictly better (storage boxes, bulk consumables) and want the best ₪-per-unit-of-stuff, not the lowest sticker.
• Downstream of search-by-synonyms / search-aliexpress on a shortlisted item.

The anomalies to catch

1. Flat price across sizes (per-volume anomaly). The classic. A "Big" variant priced ~the same as the small one. Compute ₪ per litre for every SKU; the cheapest ₪/L wins, and it's usually the largest size.
2. Steep quantity tier (per-quantity anomaly). "Buy 2+ / 5+ / 10+" drops the unit price sharply. A 10–20 unit order then lands far below 10× the single price.
3. Discount that only applies at one variant/qty. A −30/−38% badge that holds only for a size you don't want (or evaporates at the qty you'd buy). Verify at the actual size×qty you'd purchase.

Method

1. Enumerate SKUs. Open the listing in the user's own Chrome and do the locale handshake first ($CLAUDE_PLUGIN_ROOT/reference/browser.md), then read every size/variant option with its price. The size grid is JS-gated and the price XHR is anti-bot-protected, so click each option and read the updated price — plain fetch and headless both fail here. Chrome first, gateway Playwright as fallback.
2. Normalize. For each SKU, parse dimensions → volume in litres (L×W×H in cm ÷ 1000), and compute ₪/L. For non-volume goods, normalize on the natural unit (per metre, per 100 pcs, etc.).
3. Read quantity tiers. Capture any qty-break table (thresholds + unit price).
4. Flag anomalies. Rank SKUs by ₪/L (or natural unit). Call out any where a larger size is within ~10% of a smaller one's price, and any qty tier that beats the 1-unit price by a meaningful margin.
5. Landed check. Add ship-to-IL + VAT band at the winning size×qty — the anomaly only matters if it survives landed in Israel. Shipping frequently does not scale with the bulk tier, and a heavier variant can lose its free-shipping flag entirely, so re-read the lane at the winning SKU with ship-options-il rather than reusing the single-unit figure.

Output format

Listing: <title>   item <id>
Locale: ILS verified

Per-variant (sorted best ₪/L first):
  size            price     volume    ₪/L
  60×40×34 cm     ₪78.00    81.6 L    0.96   ★ cheapest per litre
  60×40×23 cm     ₪71.00    55.2 L    1.29
  40×30×15 cm     ₪69.00    18.0 L    3.83

Quantity tiers (if any):
  1+   ₪78.00   ·   5+   ₪69.00   ·   10+   ₪61.00   ← −22% vs single

Anomalies:
  • "60×40×34" priced within 10% of "60×40×23" but +48% volume → take the tall one.
  • 10+ tier −22%: a 10-unit order ≈ ₪610 vs ₪780 at single price.

Landed at winning SKU (60×40×34 ×10, ship→IL): ₪<x> incl. VAT band <…>
Beats local? local rigid 60×40 w/ lid ≈ ₪80 → compare per-box landed.

Composition

• Consumes a listing shortlisted by search-by-synonyms / search-aliexpress.
• Uses ship-options-il for the lane at the winning SKU and fetch-listing for the landed cost + VAT band.
• Feed the running total to cart-vat-nudge if buying several. A bulk tier is the most common way an otherwise-safe order crosses the $75 line — a 10-unit tier at $9/unit is $90 of goods and fully taxable. Check the tier total, not the unit price.
• find-under-75 if the goal is specifically to stay inside the exemption.

Validation checklist

1. Every visible SKU was read (not just the default), with a price each.
2. ₪/L (or the natural per-unit metric) is computed for each — not just sticker price.
3. Quantity tiers reported (including "none found").
4. The winning pick is compared landed in Israel, not on sticker alone.
5. Shipping was re-read at the winning SKU/qty, not carried over from 1 unit.
6. If a quantity tier is recommended, its total goods value was checked against the $75 de-minimis and the result stated.