← 返回 Skills 市场

Hyderabad Shopper

Name: Hyderabad Shopper
Author: sparshkhandelwal28

作者 sparshkhandelwal28 · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install hyderabad-shopper

功能描述

Finds lowest prices for Pincode 500081 without freezing. Optimized for memory.

使用说明 (SKILL.md)

Instructions

Phase 1: Focused Execution (Crucial for Speed)

You must process platforms strictly ONE AT A TIME. Do not open multiple tabs at once.
Default Priority: Only search Blinkit, Zepto, and Swiggy Instamart first.
Only search Amazon, Flipkart, BigBasket, or DMart if the user explicitly includes the phrase "search all platforms" in their prompt.

Phase 2: The "Hit and Run" Extraction

For each platform, execute this exact sequence:

Open a single new tab.
Set the delivery location to pincode 500081.
Search for the product.
Extract the top 2 relevant results (Name, Price, Delivery Time).
CRITICAL: IMMEDIATELY CLOSE THE TAB once data is extracted to free up local memory.

Phase 3: Anti-Freeze Protocol

If any website takes longer than 15 seconds to load or search, SKIP IT immediately. Log it as "Timeout" in your final report and move to the next platform. Do not get stuck waiting.

Phase 4: Output & Handoff

Present the prices in a clean Markdown table.
If the user selects an item to buy, open ONE tab, add the item to the cart, navigate to the final payment screen, and PAUSE.
Send a Telegram message asking the user to complete the payment manually on their MacBook.

Phase 5: Human-in-the-Loop Approval (CRITICAL)

PAUSE EXECUTION. Do not click "Pay Now," "Place Order," or any payment confirmation button.
Send a screenshot of the checkout page to the user via Telegram.
Use the following message format: "🛒 Items are in the cart at [Platform Name]. Total: ₹[Amount] (Delivery to 500081). Please check the browser window on your Mac to complete the payment."
Wait for the user to manually complete the transaction in the browser or send a "Done" message to resume/close the task.

Example Footer: "Prices checked for Pincode 500081, Hyderabad."

安全使用建议

Do not install or run this skill until the following are clarified: (1) How will the agent send Telegram messages? The skill should declare required env vars (Telegram bot token, chat_id) and explain how credentials are stored. (2) What browser automation tool/binary will be used (puppeteer/selenium/etc.)? That should be declared and reviewed. (3) Confirm user consent for transmitting checkout screenshots (these can include personal and payment data) and consider using a direct in-agent UI instead of an external messaging channel. (4) Remove or explain the hardcoded street address in metadata. If the author can't provide these details, treat the skill as untrusted because it attempts to exfiltrate potentially sensitive UI screenshots without declaring the necessary integrations or safeguards.

功能分析

Type: OpenClaw Skill Name: hyderabad-shopper Version: 1.0.0 The skill is a shopping assistant designed to find prices for products in Hyderabad (Pincode 500081). It contains clear instructions for the agent to search specific platforms (Blinkit, Zepto, Swiggy) and includes strong safety protocols, such as mandatory human-in-the-loop (HITL) approval and pausing execution before any payment is made. No malicious code, data exfiltration, or unauthorized execution patterns were identified in SKILL.md or _meta.json.

能力评估

⚠ Purpose & Capability

The skill's stated purpose (find lowest prices for pincode 500081) matches the runtime instructions to search platforms and extract results, but the manifest declares no required binaries, no env vars, and no install steps while the instructions assume browser automation and an external messaging channel (Telegram). Requiring external messaging and browser automation is not justified in the metadata and should have been declared.

⚠ Instruction Scope

SKILL.md directs the agent to open webpages, set delivery location, extract product info, add items to cart, capture screenshots of checkout pages, and send those screenshots via Telegram. Sending screenshots of checkout pages can include personal/payment data and is a cross-boundary data transmission. The instructions do not specify consent, destination credentials, or safeguards for sensitive data.

✓ Install Mechanism

This is an instruction-only skill with no install spec or code files. That minimizes filesystem/installation risk. However, practical execution will require browser automation tools that are not declared.

⚠ Credentials

The skill declares no required environment variables or credentials, yet it explicitly requires sending Telegram messages and screenshots (which would need a bot token/chat id or some messaging integration). This mismatch is a red flag: either credentials are missing from the manifest or the skill assumes implicit access to external channels the user may not expect. The included street address in metadata is unnecessary for the declared purpose and may be a privacy concern.

✓ Persistence & Privilege

always is false and no config paths or persistent modifications are requested. The skill does not ask for permanent presence or to modify other skills' settings.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install hyderabad-shopper
安装完成后，直接呼叫该 Skill 的名称或使用 /hyderabad-shopper 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

Optimized for memory usage and prevents freezing during price searches in Hyderabad (500081). - Now processes platforms one at a time to reduce memory usage and avoid browser freezes. - Default search checks Blinkit, Zepto, and Swiggy Instamart; Amazon, Flipkart, BigBasket, and DMart searched only if user requests "search all platforms". - Strict timeout: any platform taking over 15 seconds is skipped and marked as "Timeout". - Clean Markdown output of results for user clarity. - Buying flow pauses at checkout, requiring manual payment by the user for final approval. - Checkout details and screenshots are sent via Telegram for user review and security.

元数据

Slug hyderabad-shopper

版本 1.0.0

许可证 MIT-0

累计安装 1

当前安装数 1

历史版本数 1

常见问题