输入目标站点 + 竞品 ASIN 列表（1-20 个，整组）+ 产品图片/描述，通过统一后端扩词并生成对应站点语言的 Amazon listing。支持 US/JP/UK/DE/FR/IT/ES/CA/IN；Rufus 问答仅 US。
通过用户访问校验后，输入本地卖家精灵市场调研 Excel（近30天竞品表、近12个月竞品月表、头部ASIN关键词反查表、核心关键词转化率表），创建 market_project_<YYYYMMDD_HHmmss>/ 项目目录，在 market_research/ 下生成 Amazon 市场主报告、辅助分析 Excel、单文件 HTML 看板、运行日志和 project_manifest.json，并可在主报告后追加 Top10 品牌/卖家外部调研。用于用户已经准备好 Excel 文件、需要做市场规模、趋势、垄断度、广告压力、基础透视、规则结论、头部品牌/卖家公开信息推断分析的场景；不负责浏览器采集、商品维度打标或多维机会深挖。
输入本地西柚找词“流量得分趋势”Excel 文件夹（1-10 个竞品 ASIN，同一站点、近一年日维度），先识别和校验 ASIN、站点、日期范围、价格促销、评分评论、自然/广告流量和新增广告活动字段，再生成亚马逊竞品推广手段拆解报告；商品画像、销量预测、Woot、上架时间和类目背景由 skill 内置 CLI 自动补充。适用于分析竞品价格、促销、广告、评价、销量、Woot、样本内季节性和动作有效性；用户明确继续时，可先追加 Agent 免费公开网页证据，再由用户单独确认是否启用按实际调用计费的增强检索，并把审核后的证据合并进同一 HTML。
输入细分市场（niche）名称列表（最多 100 个，可聊天输入或从 txt/csv/xlsx 读取名称），自动从数据库拉取 9 项指标并评分排序，帮助卖家找到最容易切入的细分市场。支持 4 种卖家 profile 个性化评分。
Run a Wisdom Bud / PatSnap image-search patent workflow and write results back to the user's own Feishu/Lark Bitable. Use when Codex needs to process product rows or product images by using the user's configured Wisdom Bud account, capture candidate patent/design evidence, optionally attempt official-source verification, skip official checks that cannot be found, generate a concise report, upload it to Feishu, and update configured Feishu fields. Requires the recipient user to provide their own Wisdom Bud access and Feishu credentials/table mapping.
Run and maintain a reusable Amazon data crawler for front-end product collection with SellerSprite-enriched fields. Use when the user needs to crawl Amazon keyword search result pages with selectable sort orders, Best Sellers/New Releases category ranking nodes, storefront product lists with selectable sort orders and up to 20 pages per store, or image-search similar competitor counts/details; also use when packaging this crawler for another Codex installation.
Use when the user wants Codex to build a non-copycat Amazon image system from the user's own product photos, product facts, selling points, and either user-provided image planning ideas or a competitor Amazon listing URL/ASIN. If the user provides image generation planning or per-image design ideas, use those ideas directly and skip competitor teardown even when a competitor ASIN is also present. If only a competitor reference is provided as the strategy source, review competitor main images, gallery images, and A+ modules for strategy only. In both branches, preserve the user's real product proportions, structure, visible details, and dimension ratios from source references when available.
仅在当前对话上下文已经有 lc_amazon_market_research 刚生成并明确输出的 market_project_<YYYYMMDD_HHmmss>/ 项目根目录时，继续做 Amazon 商品机会深挖：通过服务器端关键词词根能力拆解 Top90 关键词，生成消费者决策维度候选，让 agent 完成维度确认和逐 listing 打标，再由本地 CLI 计算特征分布、维度透视、多维组合供需指数、正式机会组合和单文件 HTML 看板。上下文没有真实项目目录时必须停止，不询问路径、不扫描猜测、不重新读取原始 Excel；agent 不处理上游供应商凭据。