/parse-product-url

Drop any vendor product URL into Claude; get back a clean, structured row that fits the Canoa FF&E taxonomy. Works with Shopify product pages first (JSON shortcut), falls back to HTML extraction for vendor sites. The output is a row you can paste into a Sheet, Airtable, or your spec doc.

Why this exists

Designers spend an embarrassing share of their week copy-pasting fields out of vendor PDPs into spreadsheets. The schema is always almost-the-same, the fields are always almost-where-you-expect, and the work compounds when you’re sourcing 30+ products for a project. This skill closes the loop: any vendor URL becomes a row, in the same shape every time.

Pattern

1. Detect Shopify → use the storefront JSON (free win, structured).
2. Else → fetch the HTML, extract with a Pydantic-shaped Claude prompt.
3. Map fields onto Canoa’s schema: name, manufacturer, vendor, SKU, price, dimensions, fabric, finish, color, material, weight, lead time, sub-category, variants.
4. Return one row.

Distribution

• Claude skill — interactive single-URL or batch.
• Sheets via Apps Script — paste a list of URLs, custom function fills rows.

Status

Planned — port the existing services/openai/product_information_instructor.py to Claude tool-use. Pydantic schema already defined. Drop Playwright fallback initially (use plain HTTP + selectolax); add it back if vendor sites start blocking.