Nine agent skills for mining Amazon data through Chinese seller APIs

What it does

This project configures MCP servers for several Chinese Amazon data services—Sorftime, Sif, Xiyou, and Sellersprite—and exposes them as nine agent skills for Claude and Codex. Each skill runs a structured workflow, such as reverse-engineering a competitor listing by ASIN, grading an entire product category across five dimensions, or building an optimized product page from keyword libraries and review pain points. Outputs are packaged as Markdown reports, HTML dashboards, and structured JSON for inventory and marketing decisions.

The interesting bit

The project treats Amazon listing optimization as an AI engineering pipeline rather than mere copywriting: it assembles a “user question bank” from reviews and Reddit, gathers “evidence” for every selling point, and validates drafts against Alexa and Rufus conversational-search patterns. It also encodes fairly sophisticated e-commerce heuristics—like identifying sparse versus stable traffic periods for new products, or flagging second-hand complaint thresholds above 5%—into deterministic MCP tool chains.

Key highlights

• amazon-analyse performs a full-dimension teardown of a single ASIN, including cross-platform signals like TikTok influencer videos and 1688 sourcing costs.
• category-selection scores markets on five axes—size, growth, competition, barriers, and margin—then renders the verdict as Markdown, Excel, and HTML.
• keyword-research ingests 1,500-plus keywords via Sorftime and buckets them into eight semantic categories (material, scenario, function, etc.) to shape ad-group strategy.
• review-analysis runs a six-dimensional pain-point scan and issues automated risk warnings when service-related complaints exceed configured thresholds.
• sellersprite-amazon-research exposes 43 API tools across ten workflows, from pricing-strategy analysis to “blue ocean” opportunity mining.

Caveats

• Documentation and skill commands are presented in Chinese, so English-speaking developers will need translation to navigate the workflows.
• The project is a workflow layer atop commercial third-party data APIs; the README does not clarify authentication, pricing, or rate limits for those services.
• The provided README source is truncated, leaving the full scope of setup requirements and edge-case handling unclear.

Verdict

Worth a look if you run Amazon FBA operations and already pay for Sorftime or Sellersprite, since it turns raw API dumps into structured LLM narratives. Skip it if you are not in the Chinese Amazon seller ecosystem or lack access to the underlying data MCPs—it is purpose-built glue for that specific stack.