Birthday Cake Ordering Assistant

Problem

When a user asks to order a birthday cake through chat or automation, the task is not only product search. It also involves handling a login credential, delivery city, saved addresses, product specs, payment link generation, and order status checks. Direct API use can get messy with Chinese parameters, leaked token values, inconsistent field names, and spec mismatches.

How It Works

The skill wraps a JSON-RPC 2.0 endpoint and turns cake ordering into a repeatable workflow:

• Credential handling: the current user provides the user_token, and the skill stores it in dgss_token.md to keep sensitive data out of version control.
• Address setup: cities are normalized to the city level, then saved addresses are listed or added using the area and addr fields.
• Product search: only the cake category cat_id=1 is accepted. The result should show brand, product name, spec, price, and details.
• Order validation: before submission, the skill checks spec_id against the selected spec name, generates the payment link, and extracts the scene parameter from the returned URL.
• Order lookup: it can query the latest order status using scene, which is useful for follow-up checks or repeat purchases.

Boundaries

• It only supports cake ordering, not snacks, flowers, refunds, logistics, invoices, or similar tasks.
• One order uses a single spec and should not batch or duplicate submissions.
• On Windows, use a Python invocation path to avoid Chinese parameter encoding issues.
• User-facing output should stay concise, without exposing API domains, internal fields, or raw error traces.