Idle Fish Virtual Product Auto-Publishing Pipeline

Problem It Solves

Publishing virtual products on Xianyu by hand often means repeated login, checking base prices, choosing specs, writing titles, filling prices, uploading images, and clicking publish. This skill compresses those steps into a pipeline: after providing a Xianyu Cookie, it prices a selected product and spec, generates copy and a main image, then uses Playwright to publish.

How It Works

• Products and specs: ships with 48 virtual products and 132 specs, covering common categories such as video, cloud storage, music, and memberships.
• Dynamic pricing: reads base prices from assets/prices_cache.json, applies a 5%~10% margin to produce selling and original prices, and can preview pricing only.
• Content and image: uses DeepSeek for description copy and GPT-image-2 for the main image; because the publish page leaves the title field empty, the combined “product + spec + price + selling point” is placed in the first line of the description.
• Publishing execution: publish.py controls the workflow, deepseek_copywriter.py generates copy, and publish_one.py handles low-level browser actions; a background subprocess helps avoid interactive environment timeouts.

Boundaries

It targets automatic publishing for products already present in the spec list, not ad-hoc products, complex inventory, or compliance workflows requiring manual review. It assumes a valid Xianyu Cookie and a local browser automation environment; price caches must be updated manually, and previewing copy, image, and pricing before publishing is recommended.