kelly-pickslisted
Install: claude install-skill mr-kelly/skills
# Kelly Picks
## Overview
Use this skill as Kelly's product-research (选品) desk. The agent sweeps configured trend sources and files everything into one snapshot the local app renders:
1. **Trend feed**: raw source-tagged signals — a viral TikTok with view velocity, a BSR jump, a Temu/AliExpress riser, a rising search query, a competitor launch — each linkable to a candidate.
2. **Candidates**: products under research, each with a **margin card** (estimated price − landed cost − freight − platform fees − est. ad cost → gross margin %, breakeven ACOS) and a **competition read** (top-10 review-count distribution, head-seller dominance, new-entrant velocity).
3. **Decisions**: the review queue — the agent proposes a verdict per candidate (develop with sourcing + listing brief draft, drop with reason, keep watching with re-check criteria); Kelly approves, edits the brief, requests changes, or blocks. Approved develop items become concrete handoffs: a sourcing brief export and a listing brief for kelly-listing.
Default interaction mode: App UI. Unless the user explicitly asks for chat-only handling, check onboarding/config, refresh or load the local picks snapshot, start/reuse the local app with `app/start.sh`, and give the actual local URL. Use chat-only mode only when the user says "纯聊天", "chat only", "不要打开 UI", or similar.
Collection is agent-driven: browser automation skills in the agent session, exports, or pasted research fed through the ingest script. The app itself onl