product-picker

多平台数据加权打分筛选产品工具，针对"尚品宅配"品牌家居场景，从小红书、京东、淘宝三平台采集数据并综合评分。触发场景：(1) 选品、产品筛选；(2) 小红书/京东/淘宝数据加权打分；(3) 尚品宅配产品落地评估；(4) 产品热度/转化分析；(5) 询问"这个产品适合落地吗"类问题。输出要求：完成评估后，必须将 ....

Before installing/using this skill: 1) Confirm that your agent environment provides a trusted 'message tool' / Feishu integration and that you are comfortable the generated report will be sent to the current Feishu conversation (the skill requires this and forbids only saving locally). 2) Inspect and, if needed, edit the hard‑coded example paths and placeholders (the template shows /Users/Heaven and user:<sender_id>) so the agent won't attempt to use incorrect absolute paths or leak files to unintended recipients. 3) Verify availability and trustworthiness of the md-to-nanobanana-ppt sub-skill (the skill will call it when the user confirms PPT generation). 4) Test in a safe environment to confirm the file send behavior and that no sensitive local data is included in the generated report. 5) If you need the report to remain local only, do not install or use this skill as written — its instructions force external delivery via Feishu.

Type: OpenClaw Skill Name: product-picker Version: 1.0.1 The skill bundle defines a legitimate business SOP for multi-platform product scoring and reporting tailored for the 'Shangpin Home Collection' brand. It outlines a clear workflow for data collection from Xiaohongshu, JD, and Taobao, followed by data cleaning, weighted scoring, and report generation. The instructions to send the final Markdown report via Feishu (Lark) and call a sub-agent (md-to-nanobanana-ppt) for PPT generation are functional requirements aligned with the stated purpose. A hardcoded local path (/Users/Heaven/) in the documentation appears to be a developer artifact rather than a malicious indicator.