功能描述

Use this skill when you need real product data to answer a user's question — finding products, comparing prices, recommending items, or checking availability...

使用说明 (SKILL.md)

Product Discovery

Name: Shopping Product Search
Author: evanfen-c3

You have access to a product search script that queries a catalog of millions of products across thousands of retailers. Use it whenever you need real product data to answer a user's question.

Prerequisites

API key: This skill requires a CHANNEL3_API_KEY environment variable. Get a free key at trychannel3.com.
Dependencies: curl and jq must be installed.

Search Script

Location: product-discovery/scripts/search.sh (relative to the skill root)

Run it via the shell. The script requires curl and jq.

Usage

search.sh [OPTIONS] "query text"

Options

Flag	Description	Example
`-n NUM`	Number of results (default: 5, max: 30)	`-n 10`
`-p MAX_PRICE`	Maximum price in dollars	`-p 100`
`--min-price MIN`	Minimum price in dollars	`--min-price 50`
`-g GENDER`	Gender filter (male/female/unisex)	`-g male`
`-c CONDITION`	Product condition (new/refurbished/used)	`-c new`
`-a AGE`	Comma-separated age groups (newborn/infant/toddler/kids/adult)	`-a "kids,toddler"`
`--availability STATUS`	Comma-separated availability statuses (InStock/OutOfStock/PreOrder/BackOrder/LimitedAvailability/SoldOut/Discontinued)	`--availability "InStock"`
`-i IMAGE_URL`	Search by image (visual similarity)	`-i "https://example.com/photo.jpg"`
`-b BRAND_IDS`	Comma-separated brand IDs to include	`-b "brand_abc,brand_def"`
`-w WEBSITE_IDS`	Comma-separated website IDs to include	`-w "website_abc"`
`--categories IDS`	Comma-separated category IDs to include	`--categories "cat_abc"`
`--exclude-brands IDS`	Comma-separated brand IDs to exclude	`--exclude-brands "brand_xyz"`
`--exclude-websites IDS`	Comma-separated website IDs to exclude	`--exclude-websites "website_xyz"`
`--exclude-categories IDS`	Comma-separated category IDs to exclude	`--exclude-categories "cat_xyz"`
`--keyword-only`	Use exact keyword matching instead of semantic search	`--keyword-only`
`--next TOKEN`	Pagination token from a previous search	`--next "tok_abc..."`

The query argument is optional when using -i for image-only search. Text and image can be combined.

Examples

# Basic text search
search.sh "wireless noise cancelling headphones"

# Price-filtered search
search.sh -p 100 -n 10 "running shoes"

# Price range search
search.sh --min-price 50 -p 200 "winter boots"

# Gendered search
search.sh -g female -p 200 "winter jacket"

# Condition filter
search.sh -c used "macbook pro"

# Kids products
search.sh -a kids -p 50 "sneakers"

# Only in-stock products
search.sh --availability "InStock" "yoga mat"

# Image-based visual similarity search
search.sh -i "https://example.com/dress.jpg"

# Combined text + image search
search.sh -i "https://example.com/jacket.jpg" "similar but in blue"

# Keyword-only (exact match, no semantic search)
search.sh --keyword-only "Nike Air Max 90"

# Paginate for more results
search.sh --next "tok_abc123..." "running shoes"

Output Format

The script outputs structured text, not raw JSON. Each product includes its ID, brands, and all merchant offers with prices and buy links:

Found 5 products (next_page: tok_abc123)

1. Nike Air Zoom Pegasus 41
   ID: prod_abc123
   Brands: Nike
   Offers:
     - nordstrom.com: $89.99 (InStock) https://buy.trychannel3.com/...
     - nike.com: $94.99 (InStock) https://buy.trychannel3.com/...

2. Adidas Ultraboost Light
   ID: prod_def456
   Brands: Adidas
   Offers:
     - adidas.com: $97.00 (InStock) https://buy.trychannel3.com/...

If no results are found, the output is: No products found.

If the API key is missing or invalid, the script prints instructions on how to get one.

Workflow Patterns

Find products

User asks "find me running shoes under $100":

Run: search.sh -p 100 "running shoes"
Present the results as a clean numbered list with product name, price, merchant, and buy link.

Compare products

User asks "compare AirPods Pro vs Sony WF-1000XM5":

Run two searches: search.sh -n 3 "AirPods Pro" and search.sh -n 3 "Sony WF-1000XM5"
Build a markdown comparison table with columns for product, price, merchants, and availability.

Best option under a budget

User asks "what's the best laptop under $800":

Run: search.sh -p 800 -n 10 "laptop"
Review the results and recommend the top picks, explaining why based on the product details.

Image-based search

User shares an image URL and says "find me something like this":

Run: search.sh -i "IMAGE_URL"
Present visually similar products with prices and links.

Get more results

If the user wants more results after an initial search:

Copy the next_page token from the previous output.
Run: search.sh --next "TOKEN" "original query"
Present the additional results.

How to Present Results to the User

Synthesize the script output into a clean response. Do NOT paste the raw script output.
Present products as a numbered list or markdown table — whichever fits the question better.
Always include: product name, price, merchant name, and the buy link URL.
For comparisons, use a markdown table with columns like Product, Price, Merchant, and Link.
If multiple merchants sell the same product at different prices, highlight the cheapest option.
Keep it concise — the user wants recommendations, not a data dump.

About This Skill

This skill queries the Channel3 product catalog API (api.trychannel3.com). Search queries and any image URLs you provide are sent to this third-party API. Product buy links point to buy.trychannel3.com, which redirects to merchant sites with affiliate tracking. Avoid sending sensitive or private information in search queries.

API docs: docs.trychannel3.com
Source: github.com/channel3-ai/skills
Provider: Channel3 (trychannel3.com)

安全使用建议

This skill appears to do what it says: it runs a shell script that sends your search queries (and any image URLs you supply) to api.trychannel3.com using CHANNEL3_API_KEY and returns structured product results. Before installing: 1) Verify you trust trychannel3.com (requests and buy links go to their domains); 2) Do not include sensitive or private data in queries or image URLs because those will be transmitted to the external API; 3) Ensure curl and jq are available in your environment; 4) Keep your CHANNEL3_API_KEY secret and only grant it to this skill; and 5) Note the packaging/metadata mismatch (registry summary claiming no env vars) — confirm the skill's config requires the API key as indicated in skill.yaml/SKILL.md. If you need to be extra cautious, review the script locally (it's short and readable) before granting network/shell permissions.

功能分析

Type: OpenClaw Skill Name: channel3-product-discovery Version: 1.0.1 The skill provides a legitimate interface for searching products via the Channel3 API. The bash script (scripts/search.sh) safely constructs JSON requests using jq and communicates with a known endpoint (api.trychannel3.com). It requires standard permissions (network, shell) and handles the API key as a secret. No indicators of data exfiltration or malicious execution were found.

能力评估

ℹ Purpose & Capability

The skill's name/description, skill.yaml, SKILL.md, and the shell script all align: this is a product search that calls Channel3's API and returns merchant links. However, the top-level registry summary in the evaluation metadata (which said "Required env vars: none") conflicts with skill.yaml and SKILL.md, both of which require a CHANNEL3_API_KEY and list curl/jq as dependencies. This appears to be a packaging/metadata inconsistency rather than malicious misdirection.

ℹ Instruction Scope

The runtime instructions and script only build a JSON body from CLI args and POST it to https://api.trychannel3.com/v1/search using CHANNEL3_API_KEY; they do not read arbitrary files or other env vars. Important: user queries and any provided image URLs are transmitted to the third-party API (expected for this purpose). If a user includes sensitive data in the query or image URL, that data will be sent to the external service.

✓ Install Mechanism

No install spec; it's instruction-only with a bundled shell script. This is low-risk: nothing is downloaded or written during install. The script expects curl and jq to be present (which SKILL.md documents).

ℹ Credentials

The only secret required is CHANNEL3_API_KEY (declared in skill.yaml and SKILL.md, marked secret). That is proportionate to calling a third-party product API. The metadata inconsistency (some summary saying no env vars required) should be fixed so users aren't misled.

✓ Persistence & Privilege

The skill is not always:true, does not request elevated/persistent system changes, and does not modify other skills. It requires network and shell permissions which are appropriate for executing the provided script and making API requests.

版本历史

v1.0.1

**channel3-product-discovery 1.0.1 Changelog** - Added a skill.yaml file for proper skill packaging and metadata. - Updated documentation to clarify API key requirements, dependencies (`curl` and `jq`), and provider information. - Added an "About This Skill" section detailing API usage and privacy notes. - No changes to core functionality; enhancements are documentation and configuration only.

v1.0.0

Initial release of the product-discovery skill. - Enables real-time product searches, comparisons, price checks, and recommendations across millions of products and thousands of retailers. - Provides a search script with extensive filtering options (price, gender, age, brand, condition, availability, etc.). - Supports both text and image-based searches, including combined queries. - Outputs structured product results with merchant pricing and buy links for easy presentation. - Includes detailed usage patterns and workflow recommendations for shopping, price comparison, and product discovery tasks.

元数据

Slug channel3-product-discovery

版本 1.0.1

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 2

常见问题

Shopping Product Search 是什么？

Use this skill when you need real product data to answer a user's question — finding products, comparing prices, recommending items, or checking availability... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 246 次。

如何安装 Shopping Product Search？

在 OpenClaw 或 Claude Code 对话框中运行命令「/install channel3-product-discovery」即可一键安装，无需额外配置。

Shopping Product Search 是免费的吗？

是的，Shopping Product Search 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。

Shopping Product Search 支持哪些平台？

Shopping Product Search 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。

谁开发了 Shopping Product Search？

由 evanfen-c3（@evanfen-c3）开发并维护，当前版本 v1.0.1。

Shopping Product Search