Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Amazon Deep Research V4

v1.0.0-beta.1

亚马逊深度选品V4:多平台交叉验证(AMZScout+西柚找词+卖家精灵) → 1688以图搜图同款采购 → 实时汇率7项毛利计算 → WIPO IP合规审计 → HTML可视化报告。 触发词:亚马逊调研/选品/市场调研/Amazon research/毛利分析/快速调研/产品分析/竞品分析/选品报告

0· 138·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for zachary2024/amazon-deep-research-v4.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Amazon Deep Research V4" (zachary2024/amazon-deep-research-v4) from ClawHub.
Skill page: https://clawhub.ai/zachary2024/amazon-deep-research-v4
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install amazon-deep-research-v4

ClawHub CLI

Package manager switcher

npx clawhub@latest install amazon-deep-research-v4
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
Name/description (multi-platform Amazon research + 1688 sourcing + IP audit + HTML report) align with the sites and data the SKILL.md instructs the agent to access (AMZScout, 西柚找词, 卖家精灵, Amazon, 1688, WIPO, Google Patents). The requested actions are coherent for the stated product-research workflow.
!
Instruction Scope
The runtime instructions require extensive automated browser scraping of multiple third-party sites and a web_search for FX rates (expected). However, the skill explicitly instructs using the browser to extract cookies via document.cookie (for 卖家精灵) which is sensitive: document.cookie can expose other site cookies if mishandled. The SKILL.md does not constrain how cookies are collected, stored, or limited to the few named keys, nor does it instruct safe handling/expiry or require explicit user consent flows. It also requires extracting data from many external pages which could capture additional private data (e.g., session metadata) if implemented broadly.
Install Mechanism
Instruction-only skill with no install spec or packages — lowest install risk. No downloads or archive extraction are present.
!
Credentials
The skill declares no required env vars or credentials, which superficially lowers risk. But it asks the agent to obtain session cookies (ecookie, rank-login-user, Sprite-X-Token) from the user's browser after login. Requesting browser cookies is a form of credential access; the SKILL.md does not limit extraction to those three keys or explain secure handling, retention, or scope. This raises proportionality and privacy concerns even though the cookies are relevant to calling 卖家精灵's API.
Persistence & Privilege
The skill does not request persistent/always-on presence (always:false) and contains no instructions to modify other skills or system-wide settings. Autonomous invocation is allowed (platform default) but not combined with other high privileges.
What to consider before installing
What to consider before installing: 1) Cookie extraction: the skill tells the agent to extract browser cookies (document.cookie) after you log into 卖家精灵 — that can expose other cookies if implemented carelessly. Only proceed if you trust the skill author and you understand how those cookies will be provided/stored and that they will be limited to the listed keys. 2) Scraping and TOS: the skill automates scraping of AMZScout, Amazon, 1688, 西柚找词 and others — this can violate those sites' terms and may trigger blocking or account action. 3) IP/legal checks: the WIPO/ patent checks are non-trivial legal tasks; a human IP lawyer review is recommended for borderline results. 4) No install reduces supply-chain risk, but the skill can still exfiltrate browser data via the instructed browser agent. 5) Recommendations: ask the publisher for source code or an implementation plan showing cookie handling and storage policies; require explicit prompts/consent before any cookie capture; restrict cookie extraction to the named keys and auto-delete them after use; run the skill on test accounts first; and verify compliance with target sites' terms. If the author cannot clearly explain cookie handling and data retention, treat the skill as risky.

Like a lobster shell, security has layers — review code before you run it.

latestvk97eg67aw24ytk6cyeqkgvr5rx84hv82
138downloads
0stars
1versions
Updated 2w ago
v1.0.0-beta.1
MIT-0

亚马逊深度选品 V4 (多平台交叉验证版)

核心原则

  1. 所有数据必须真实抓取,严禁估算或编造
  2. 多平台交叉验证:AMZScout(费用) + 西柚找词(流量) + 卖家精灵(BSR/维度) + Amazon官方(尺寸/重量)
  3. 1688以图搜图:用Amazon主图精准匹配同款,不接受关键词模糊搜索
  4. 实时汇率:每次执行必须 web_search 获取当日 USD/CNY 汇率
  5. WIPO IP 合规:每个产品必须完成5步IP审计,HIGH风险直接淘汰
  6. 毛利7项扣除不可省略,≥20%合格,<20%淘汰

流程(6步)

#步骤数据来源输出
1获取实时汇率web_search Yahoo FinanceFX 变量
2多平台数据采集AMZScout + 西柚找词 + 卖家精灵data/multi_platform.json
31688以图搜图采购匹配air.1688.com 图片搜索data/1688_matches.json
47项毛利计算+筛选Pythondata/qualified.json
5WIPO IP 合规审计WIPO Brand DB + Google Patentsdata/ip_audit.json
6HTML报告生成Python(内嵌模板)output/amazon-research-v4.html

执行指令

步骤1:实时汇率

# 必须每次执行时获取,不使用固定值
web_search("USD CNY exchange rate today")
# 提取当日汇率,如 6.8335
FX = 提取到的实时汇率

步骤2:多平台数据采集

2a. AMZScout(费用权威来源)

用 browser agent 访问 AMZScout Product Database:

https://amzscout.net/product-database
  • 输入ASIN搜索或按品类筛选
  • 提取字段:price, fba_fee(含Referral), monthly_sales, weight, bsr, rating, reviews
  • 关键:AMZScout 的 Fees 字段 = FBA履约费 + 15% Referral 佣金合计,计算毛利时不再重复扣佣金

2b. 西柚找词(流量权威来源)

用 browser agent 逐ASIN查询:

https://www.xiyouzhaoci.com/detail/asin/look_up/US/{ASIN}
  • 提取字段:
    • monthly_orders_30d (30天预估单量)
    • traffic_score_7d (7天流量分)
    • natural_traffic_pct (自然流量占比%)
    • ad_traffic_pct (广告流量占比%)
    • top_keywords: keyword, search_volume, natural_rank, ad_rank, cpc (前3-5个)
  • 关键:西柚的广告占比数据是广告投放策略的核心依据

2c. 卖家精灵(BSR/维度备选)

登录方式:

browser → https://www.sellersprite.com/cn/w/user/login
用户扫码登录后提取Cookie: ecookie, rank-login-user, Sprite-X-Token

API采集(参照 references/sellersprite-api.md):

POST https://www.sellersprite.com/v3/api/product-research

Extension详情页(单ASIN):

https://www.sellersprite.com/cn/extension/product-detail?market=US&asin={ASIN}
  • 提取:fba, dimensions, dimensionsTag, weight
  • 注意:卖家精灵的FBA数据可能不如AMZScout准确,优先用AMZScout

2d. Amazon官方(尺寸/重量地面真相)

browser agent 访问 Amazon 产品页:

https://www.amazon.com/dp/{ASIN}
  • 滚动到 "Product information" / "Technical Details"
  • 提取:Product Dimensions, Item Weight, Package Weight, Date First Available
  • Amazon官方数据是尺寸/重量的最终权威

交叉验证规则

数据项第一来源验证来源冲突处理
FBA+佣金费用AMZScout卖家精灵取AMZScout(含Referral更完整)
月销量AMZScout西柚找词取两者平均值
尺寸/重量Amazon官方卖家精灵取Amazon官方
流量/广告占比西柚找词唯一来源
关键词SV/CPC西柚找词唯一来源
BSRAMZScout卖家精灵取更新时间较近者

步骤3:1688以图搜图

必须用Amazon主图做图片搜索,不接受关键词搜索(产品不匹配)

搜索URL:

https://air.1688.com/app/1688-lp/landing-page/comparison-table.html?bizType=browser&currency=CNY&customerId=dingtalk&outImageAddress={URL_ENCODED_AMAZON_IMAGE}

执行方式:browser agent 逐ASIN处理

  1. 获取Amazon主图URL(#landingImage src)
  2. URL编码主图地址
  3. 导航到 air.1688.com 搜索URL
  4. 等待5秒加载结果
  5. 提取TOP匹配:产品名、价格(¥)、detail.1688.com链接、供应商名

提取JS:

const rows = document.querySelectorAll('tr.ant-table-row');
const results = [];
rows.forEach((row, i) => {
  if (i >= 3) return;
  const tds = row.querySelectorAll('td');
  const img = row.querySelector('img')?.src || '';
  const name = tds[1]?.innerText || '';
  const price = tds[2]?.innerText || '';
  results.push({img, name, price});
});
JSON.stringify(results);

约束

  • 链接必须是 detail.1688.com/offer/ 格式
  • 不接受 alibaba.com 国际站链接
  • 匹配度标注:✓(高度匹配) / ~(近似匹配) / ✗(不匹配,需人工确认)

步骤4:7项毛利计算

参照 references/margin.md

毛利 = 售价 - ①采购成本 - ②头程运费 - ③关税 - ④AMZScout总费 - ⑤退货 - ⑥广告
#项目公式来源
采购成本1688价(¥) ÷ 实时汇率1688以图搜图
头程运费重量(kg) × ¥15/kg ÷ 实时汇率Amazon官方重量
关税采购成本($) × 关税率(5%-10%)品类查表
AMZScout总费AMZScout Fees (FBA+佣金)AMZScout
退货售价 × 5%固定
广告售价 × 20%固定(TACOS)

注意:AMZScout Fees 已包含 FBA 履约费 + 15% Referral 佣金,不再单独扣佣金。

筛选:≥20% 合格 | <20% 淘汰 色标:≥35% 绿色优秀 | ≥20% 黄色合格 | <20% 红色淘汰

步骤5:WIPO IP 合规审计

每个产品必须完成,不可跳过(参照 references/wipo-ip.md

5步审计流程:

  1. 商标检索 → WIPO Global Brand Database (branddb.wipo.int)
  2. 外观专利 → Google Patents (D前缀, <15年)
  3. 实用新型专利 → WIPO PATENTSCOPE + USPTO
  4. 认证合规 → DOT/FCC/UL/CARB/FMVSS/EPA/MFi
  5. 风险评级 → 🟢LOW / 🟡MEDIUM / 🔴HIGH

HIGH 风险产品直接淘汰,不进入最终报告

步骤6:HTML报告生成

参照 references/html-template.md

报告结构:

标题 + 数据来源标注(AMZScout/西柚/1688)
常量栏(实时汇率/运费/佣金/退货/广告/FBA来源)
7项毛利公式框(黑底白字)
KPI卡片(合格数/平均毛利/费用来源/流量来源)
产品明细表(含全部7项扣除拆解)
数据来源说明

产品表列: #, 主图(Amazon), ASIN(链接), 产品名, AMZScout数据(月销/BSR/评分/FBA/佣金), 西柚数据(流量分/自然占比/广告占比), 核心关键词(SV/CPC), 售价, 1688采购(¥/链接), 供应商, ①成本, ②运费, ③关税, ④总费, ⑤退货, ⑥广告, 净利, 毛利率(色标), IP风险

核心约束

  1. 数据必须多平台实采,禁止单一来源估算
  2. FBA费用取AMZScout,不用公式估算
  3. 1688必须以图搜图,不接受关键词搜索(产品不匹配)
  4. 1688链接必须是 detail.1688.com/offer/
  5. 汇率必须实时获取(web_search),不用固定值
  6. WIPO IP必须审计,HIGH风险直接淘汰
  7. 毛利7项扣除不可省略
  8. 过滤大品牌和高退货品类(参照 references/filters.md
  9. 报告中每个数字必须标注数据来源
  10. 主图必须从Amazon真实抓取,不接受占位符

Comments

Loading comments...