Imported from topviewai/skill (
topview-shopee-ops/SKILL.md). Install upstream withnpx skills add topviewai/skill --skill topview-shopee-ops. Copyright stays with the author.
Topview Shopee Ops
Public Shopee intelligence via the host's unified topview-mcp server (progressive disclosure; domain shopee; deferred queryshopee* tools).
Step 0 — Bootstrap
- Confirm MCP server
topview-mcpis connected (look foruse_shopee_data,get_tool_schema,call_tool). - If missing, point the user to
mcp.example.json→~/.cursor/mcp.json, reload MCP. - Do not run
scripts/mcp_call.py, open SSE yourself, or use curl/raw HTTP.
Execution
Server: topview-mcp (unified). Never open SSE yourself.
Shopee data tools are deferred. Invoke via progressive disclosure:
- Read the
use_shopee_datadescription ontools/list(embedded Tools catalog). Callinguse_shopee_dataonly to list tools is optional; call it once if the catalog says it is not loaded yet (Shopee proxy warm-up). - Pick a deferred
queryshopee*name from the catalog or references/tool_routing.md. get_tool_schemawith{ "name": "<deferred>", "domain": "shopee" }.call_toolwith{ "name": "<deferred>", "domain": "shopee", "arguments": { ... } }.
Typical host tool ids: mcp__topview-mcp__get_tool_schema / mcp__topview-mcp__call_tool.
Resolve category/item/shop/brand IDs before batch detail. Always confirm site + date window when required.
UX Rules
- Be concise; lead with findings.
- Ask for site / category / ids when missing.
- Do not invent private shop ads/order/profit data.
- Creative production →
topview-generate. - Match the user's language.
Source
Adapted from Marketing Studio Shopee expert skills (V060/V061).