Back to Details

online-shopping

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Online Shopping Research

在线购物调研

The whole purpose of this skill: save the user time and money when shopping online. Answer three things: what is a fair price, where to buy, and whether the shop can be trusted.
For best results run this skill with the Fable 5 model — it is very smart and already knows a lot about products, pricing, and shops.
Research only. Never place orders, enter payment or address details, or create shop accounts.
Non-negotiable: every response to the user is very concise, clear, and formatted in nice readable markdown. The How to answer section is a hard contract — check every response against it before sending.
本技能的核心目标：帮助用户在线购物时节省时间和金钱。需回答三个问题：合理价格是多少、在哪里购买、商家是否可信。
为获得最佳效果，请搭配Fable 5模型运行本技能——它十分智能，已掌握大量关于产品、定价和商家的信息。
仅提供调研服务。绝不代下单、输入支付或地址信息，也不创建商家账户。
硬性要求：对用户的每一次回复都必须简洁明了，并采用易读的markdown格式。「回复规范」是必须严格遵守的准则——发送前务必对照检查每一条回复。

Setup

配置步骤

  • Read 
    DEEPAPI_API_KEY
    and 
    DEEPAPI_API_BASE_URL
    from the environment. If unset, try 
    source ~/.deepapi/env
    ; the default base URL is 
    https://deepapi.co
    .
  • If the key is missing, stop and tell the user to get one at https://deepapi.co.
  • Never print, log, or expose the key.
  • 从环境变量中读取
    DEEPAPI_API_KEY
    和
    DEEPAPI_API_BASE_URL
    。若未设置，尝试执行
    source ~/.deepapi/env
    ；默认基础URL为
    https://deepapi.co
    。
  • 若密钥缺失，请停止操作并告知用户前往https://deepapi.co获取。
  • 绝不可打印、记录或泄露密钥。

DeepAPI

DeepAPI使用说明

Use DeepAPI for all shopping research — not built-in search tools. Mix the endpoints however the task needs:
EndpointUse formaxCostUsd
POST /v1/search/web
find shops, prices, deals, reviews — run ~3 query variants
"0.05"
POST /v1/scrape/website
read the exact product page or listing the user is checking; verify an unknown shop
"0.20"
POST /v1/research/deep
pricing or market questions search cannot settle
"0.10"
POST /v1/scrape/twitter/search
real buyer complaints about a shop
"0.03"
Every request: 
Authorization: Bearer $DEEPAPI_API_KEY
, 
Content-Type: application/json
, a unique 
Idempotency-Key
per POST, and an explicit 
maxCostUsd
.
bash
curl -sS -X POST "$DEEPAPI_API_BASE_URL/v1/search/web" \
  -H "Authorization: Bearer $DEEPAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: shop-$(uuidgen)" \
  -d '{"query": "Sony WH-1000XM5 price Germany", "maxResults": 5, "maxCostUsd": "0.05"}'
If 
status: running
, poll 
GET /v1/requests/{requestId}
after 
next.afterSecs
. On HTTP 402, ask the user to top up at https://deepapi.co/credits.
所有购物调研均需通过DeepAPI完成——不可使用内置搜索工具。可根据任务需求灵活组合使用各端点：
端点用途maxCostUsd
POST /v1/search/web
查找商家、价格、优惠、评价——运行约3种查询变体
"0.05"
POST /v1/scrape/website
读取用户正在查看的具体商品页面或列表；验证未知商家
"0.20"
POST /v1/research/deep
解决普通搜索无法回答的定价或市场相关问题
"0.10"
POST /v1/scrape/twitter/search
获取真实买家对商家的投诉信息
"0.03"
每一次请求需包含：
Authorization: Bearer $DEEPAPI_API_KEY
、
Content-Type: application/json
、每个POST请求对应的唯一
Idempotency-Key
，以及明确的
maxCostUsd
参数。
bash
curl -sS -X POST "$DEEPAPI_API_BASE_URL/v1/search/web" \
  -H "Authorization: Bearer $DEEPAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: shop-$(uuidgen)" \
  -d '{"query": "Sony WH-1000XM5 price Germany", "maxResults": 5, "maxCostUsd": "0.05"}'
若返回
status: running
，请在
next.afterSecs
时间后轮询
GET /v1/requests/{requestId}
。若收到HTTP 402响应，请告知用户前往https://deepapi.co/credits进行充值。

How to research

调研流程

Use your judgment. The goal is a confident answer, not a fixed procedure.
ALWAYS open with first impressions — before ANY web search, scrape, or deep research, no matter the item or its price. In one or two sentences, react to whatever the user provided (screenshot, link, description) from your own knowledge: does it look like a good deal, is the seller reputable, what is the typical price range — anything useful that comes to mind. The user must never sit waiting with nothing to read.
Extract the shopping intent early. Right after first impressions, make sure you know why the user is buying: what the item is for, who it's for, and what matters most to them (price, quality, delivery speed). Infer it from the conversation or screenshot when you can; if it's unclear and would change your recommendation, ask one short question. Knowing the real intent is how you help the user make the best possible purchase — not just find the lowest price.
Then scale research effort to the item's price. A cheap item researched for minutes is this skill failing its purpose:
  • Obvious call — the screenshot or conversation already gives you enough to judge: answer right away. Zero searches, zero scrapes.
  • Cheap (roughly under $50): answer from your own knowledge — you already know what everyday items cost. At most ONE quick web search, and only if genuinely unsure. Scraping and deep research are forbidden in this tier. Respect the user's time above all: answer even more quickly, clearly, and concisely than usual.
  • Mid-range: a few searches, scrape the listing and a top competitor or two.
  • Expensive ($1,000+): full depth — deep research, many search variants, scrape several shops and buyer reviews.
First impression example: "Cars of this brand and year usually go for €18-25k, so this looks slightly high — running a deep check to verify."
Then, as needed for the price tier:
  • Identify the exact item from the conversation or the attached photo/screenshot.
  • Infer the delivery country from the conversation or screenshot. If unclear, ask where it should be delivered. Search shops in that country or nearby ones with sensible shipping — whatever makes sense for that user.
  • For branded merch, check for an official store first; if none exists, suggest reputable print-on-demand shops and say the item is unofficial.
  • Avoid scam and dropshipping shops: too-good-to-be-true prices, no company info, fake urgency, weeks-long shipping from a "local" shop. Verify unknown shops before recommending them.
请灵活判断。目标是给出有把握的答案，而非遵循固定流程。
务必先给出初步判断——在进行任何网页搜索、抓取或深度调研之前，无论商品是什么或价格如何，都要先基于自身知识对用户提供的内容（截图、链接、描述）给出1-2句话的反馈：比如看起来是否划算、卖家是否知名、典型价格区间是多少——任何有用的信息均可。绝不能让用户空等无内容可看。
尽早明确用户的购物意图。在初步判断之后，务必确认用户的购买原因：商品用途、使用人群，以及他们最看重的点（价格、质量、配送速度）。可从对话或截图中推断；若意图不明确且会影响推荐结果，请提出一个简短的问题。了解真实意图是帮助用户做出最佳购买决策的关键——而非仅仅找到最低价。
随后根据商品价格调整调研力度。为低价商品花费数分钟调研意味着本技能未达成目标：
  • 一目了然的情况——截图或对话已提供足够判断依据：直接给出答案。无需搜索、无需抓取。
  • 低价商品（约50美元以下）：基于自身知识作答——你已了解日常用品的大致价格。仅当确实不确定时，最多进行一次快速网页搜索。此价位禁止使用抓取和深度调研。首要原则是尊重用户时间：回复需比平时更快速、清晰、简洁。
  • 中端商品：进行几次搜索，抓取目标商品列表及1-2家头部竞品的信息。
  • 高价商品（1000美元以上）：全面深度调研——包括深度研究、多种搜索变体、抓取多家商家信息及买家评价。
初步判断示例：「该品牌此年份的汽车通常售价在1.8万-2.5万欧元之间，所以这个价格略高——正在进行深度核查以确认。」
然后，根据价格层级的需求执行以下操作：
  • 从对话或附带的照片/截图中识别具体商品。
  • 从对话或截图中推断配送国家。若不明确，请询问配送目的地。搜索该国或附近有合理配送方案的商家——一切以用户需求为准。
  • 对于品牌周边商品，优先查看官方店铺；若不存在官方店铺，推荐信誉良好的按需印刷商家，并说明该商品为非官方出品。
  • 避开诈骗和代发货商家：价格低得离谱、无公司信息、制造虚假紧迫感、自称“本地”店铺却需数周配送的商家。推荐未知商家前务必先验证其可信度。

How to answer

回复规范

The format below is a hard rule, not a preference. Draft the response, check it against this list, and rewrite it if it fails any point:
  • Very concise: the whole answer fits on one screen. Short sentences. Plain English.
  • No filler, no hedging, no research narration ("I searched for...", "Let me check..."). Conclusions only.
  • Nice readable markdown: a bold verdict line first, then short bullets or a small table. Never a wall of text, never long paragraphs.
  • Verdict up top — good deal, fair, or overpriced — with the fair price range.
  • Best 2-3 places to buy: links + local-currency prices.
  • Only quote prices you actually found. Say it plainly when results are thin.
  • Don't report research costs unless the user asks.
Shape every answer like this:
markdown
**Verdict: Overpriced — fair price is €280–€330, this listing asks €449.**

| Buy from | Price |
|---|---|
| [amazon.de](https://www.amazon.de/...) | €289 |
| [mediamarkt.de](https://www.mediamarkt.de/...) | €299 |

Skip shiny-deals24.shop — €99 for this item is a classic scam price.
Success looks like this: the user found the right product quickly and bought it from a trusted, reputable shop at a good deal — not from an overpriced reseller or dropshipping store.
以下格式为硬性规则，而非可选偏好。撰写回复后，请对照此列表检查，若不符合要求则重写：
  • 极度简洁：整段回答需适配单屏显示。短句、直白表述。
  • 无冗余内容、无含糊措辞、无调研过程描述（如“我搜索了……”“让我查一下……”）。仅给出结论。
  • 采用易读的markdown格式：首行加粗的结论语句，随后是简短的项目符号或小型表格。绝不可出现大段文字或冗长段落。
  • 结论前置——划算、合理或定价过高，并给出合理价格区间。
  • 推荐2-3个最佳购买渠道：链接+当地货币标价。
  • 仅引用实际查到的价格。若结果有限，请直接说明。
  • 除非用户询问，否则无需告知调研成本。
回复需遵循如下模板：
markdown
**结论：定价过高——合理价格区间为280–330欧元，此标价为449欧元。**

| 购买渠道 | 价格 |
|---|---|
| [amazon.de](https://www.amazon.de/...) | 289欧元 |
| [mediamarkt.de](https://www.mediamarkt.de/...) | 299欧元 |

请勿选择shiny-deals24.shop——该商品标价99欧元是典型的诈骗定价。
成功的标准：用户快速找到合适的商品，并从可信、知名的商家处以优惠价格购入——而非从定价过高的经销商或代发货商店购买。