Content
# Tool List
Based on SellerSprite MCP services and Agent Skills, this is a set of Amazon research tools covering competitor listing analysis, category selection, keyword research, comment VOC analysis, product selection in-depth research, and product planning.
> This repository is a desensitized open-source template that does not contain real API keys, real product reports, seller data, ASIN data, comment text, or private product selection conclusions. All keys are configured through local `.mcp.json` or `.env`.
SellerSprite MCP official documentation: https://open.sellersprite.com/mcp
## Project Introduction
This project configures the SellerSprite cross-border e-commerce data service MCP server and provides six core skills:
| Skill | Analysis Object | Command | Purpose |
|---|---|---|---|
| `amazon-analyse` | Single Listing | `/amazon-analyse {ASIN} {SITE}` | Competitor listing full-dimensional penetration analysis |
| `category-selection` | Entire category | `/category-select "{category}" {SITE}` | Category automated product selection analysis |
| `keyword-research` | Keyword library | `/keyword-research {ASIN} {SITE}` | Keyword in-depth research and 8-dimensional intelligent classification |
| `review-analysis` | User reviews | `/review-analysis {ASIN} {SITE}` | Comment in-depth analysis and pain point mining |
| `product-research` | Product selection in-depth research | `/product-research "{product keyword}" {SITE}` | LLM-driven product selection in-depth research and decision-making |
| `product-planning` | Product planning | `/product-planning "{product planning}" {SITE}` | V1 standard planning: intended product, market, competitor optimization, SWOT, ABA, MCP source data, and data source |
## Core Functionality
### Listing Level Analysis `amazon-analyse`
- **Competitor Listing Analysis**: Automatically obtain product details, reviews, keywords, and sales trend data.
- **Keyword Analysis**: Analyze traffic sources, competitor keyword layout, and long-tail word opportunities.
- **Comment Sentiment and VOC Analysis**: Cluster user likes, negative reviews, and improvement suggestions.
- **Data Verification**: Add Gemini/GLM/rule verification results at the end of the report.
- **Excel Meeting Version**: After saving the Markdown report, optionally convert it to a product manager meeting version Excel, and retain the `MCP Raw Data` page for tracing.
### Category Level Analysis `category-selection`
- **Market Analysis**: Top products, sales, sales revenue, price band, number of reviews, and ratings.
- **Five-Dimensional Scoring Model**: Market size, growth potential, competition intensity, entry barriers, and profit space.
- **Opportunity Identification**: Identify low-competition price bands, comment pain points, keyword gaps, and entry angles.
- **Entry Suggestions**: Output entry, observation, or abandonment suggestions with evidence chains.
### Keyword In-Depth Research `keyword-research`
- **Keyword Library Collection**: Obtain ASIN traffic words and related extension words through SellerSprite MCP.
- **8-Dimensional Intelligent Classification**: Negative words, brand words, material words, scenario words, attribute words, functional words, core words, and others.
- **Advertising Strategy Guidance**: Negative word list, precise matching group, scenario ad group, and broad matching group.
- **Multi-Format Output Suggestions**: Markdown report, CSV keyword library, negative word list, and classified statistics JSON.
### Comment In-Depth Analysis `review-analysis`
- **Positive and Negative Comment Layering**: Analyze 4-5 star positive feedback and 1-3 star negative feedback separately.
- **Pain Point Clustering**: Product defects, design functional defects, material appearance issues, description inconsistencies, and service logistics issues.
- **Risk Warning**: Identify high return risk, high negative review triggers, and after-sales risk.
- **Dual-Track Solution**: Product improvement suggestions + Listing/customer service script optimization suggestions.
### Product Selection In-Depth Research `product-research`
- **LLM-Driven Analysis**: Conduct market, competitor, VOC, barrier, and opportunity judgments around product keywords.
- **Complete Process**: Information collection → Data collection → Attribute labeling → Cross-analysis → Competitor and VOC → Assessment decision → Report output.
- **Decision Output**: Enter, observe, or abandon with verification indicators and next steps.
- **Multi-Format Output Suggestions**: Markdown complete report, structured data, and visual dashboard.
### Product Planning `product-planning`
- **V1 Planning Closed Loop**: Intended product → Market analysis → Competitor analysis and optimization strategy → SWOT → ABA/keywords → MCP raw data → Data source.
- **Evidence-Driven**: Planning conclusions must come from SellerSprite MCP, Amazon front-end, existing research reports, user input, or explicitly labeled external supplementary data.
- **Meeting Version Excel**: Do not output `cost calculation` and `sales plan` by default; if financial calculation is required, users must provide cost, logistics, selling price, advertising ratio, and other assumptions before adding separately.
- **Traceable Output**: Each key conclusion must be traceable to MCP tools, input summary, return fields, or data gaps.
## Quick Start
### Environmental Requirements
- Claude Code CLI, Codex, or other environments compatible with Agent Skills
- Valid SellerSprite MCP API Key
- Python 3.10+ (recommended)
- Gemini API Key, GLM API Key: used for complete dual-model cross-validation; if missing, the runtime will actively ask, and cannot claim to have completed real Gemini/GLM verification.
### Obtain SellerSprite API Key
1. Log in to SellerSprite.
2. Open the SellerSprite MCP/API related page.
3. Create or copy your MCP API Key.
4. Do not submit real keys to GitHub.
This repository only provides placeholder examples:
```json
{
"mcpServers": {
"sellersprite": {
"type": "streamableHttp",
"url": "https://mcp.sellersprite.com/mcp",
"headers": {
"secret-key": "${SELLERSPRITE_API_KEY}"
},
"name": "SellerSprite MCP"
}
}
}
```
### Configure MCP
Copy example files:
```bash
cp .mcp.json.example .mcp.json
cp .env.example .env
```
Edit `.mcp.json`:
```json
{
"mcpServers": {
"sellersprite": {
"type": "streamableHttp",
"url": "https://mcp.sellersprite.com/mcp",
"headers": {
"secret-key": "${SELLERSPRITE_API_KEY}"
},
"name": "SellerSprite MCP"
}
}
}
```
Edit `.env`:
```text
SELLERSPRITE_API_KEY=YOUR_SELLERSPRITE_API_KEY
SELLERSPRITE_MCP_URL=https://mcp.sellersprite.com/mcp
GEMINI_API_KEY=YOUR_GEMINI_API_KEY
GLM_API_KEY=YOUR_GLM_API_KEY
GLM_BASE_URL=https://open.bigmodel.cn/api/paas/v4
DEFAULT_AMAZON_SITE=US
REPORTS_DIR=reports
```
`.mcp.json` and `.env` are ignored in `.gitignore`, please do not submit.
### Runtime Credential Pre-Check
The open-source version will not carry any real keys. Each skill must perform credential pre-check before formally collecting data:
| Credential | Purpose | Behavior if missing |
|---|---|---|
| `SELLERSPRITE_API_KEY` | Call SellerSprite MCP to obtain Amazon market, competitor, keyword, and review data | Actively ask users; stop data-type reports if not provided |
| `GEMINI_API_KEY` | Gemini as cross-validation judge | Actively ask users; cannot write "executed Gemini verification" if not provided |
| `GLM_API_KEY` | GLM as cross-validation judge | Actively ask users; cannot write "executed GLM verification" if not provided |
You can run locally:
```bash
python scripts/python/sellersprite_skill_tools.py check-runtime-config --strict
```
If Windows does not add Python to PATH, you can use the actual interpreter path, for example:
```text
%LOCALAPPDATA%\Python\pythoncore-3.14-64\python.exe scripts/python/sellersprite_skill_tools.py check-runtime-config --strict
```
This script only shows whether the configuration is complete and does not print keys.
If credentials are missing during runtime, the Agent should prompt:
```text
The current process requires real data and dual-model verification, but the following credentials are missing:
- SellerSprite MCP API key: used to call SellerSprite MCP to obtain Amazon market/competitor/keywords/review data
- Gemini API key: used for Gemini data verification
- GLM API key: used for GLM data verification
Please provide the missing key, or configure it in local `.env` / `.mcp.json` and let me continue. After receiving it, I will only use it for local running and will not write it into the report or open-source directory.
```
### Install Skills
Claude Code style:
```bash
python scripts/python/sellersprite_skill_tools.py install-skill --target claude
```
Codex style:
```bash
python scripts/python/sellersprite_skill_tools.py install-skill --target codex
```
Install a single skill:
```bash
python scripts/python/sellersprite_skill_tools.py install-skill --target codex --skill amazon-analyse
```
## Using Analysis Skills
### 1. Listing Competitor Analysis
```text
# Analyze US station competitors
/amazon-analyse {ASIN} US
# Analyze German station competitors
/amazon-analyse {ASIN} DE
```
Report saved:
```text
reports/{ASIN}_{SITE}_{YYYYMMDD}/
├── report.md
├── raw/
│ ├── asin_detail.clean.json
│ ├── asin_sales_trend.clean.json
│ ├── traffic_keyword.clean.json
│ ├── traffic_extend.clean.json
│ └── reviews_*.clean.json
└── optional_excel.xlsx
```
Markdown report generation will prompt:
```text
Do you need to convert to Excel table output?
```
Regardless of whether to convert to Excel, the Markdown report must be retained.
If you choose to convert to Excel, the workbook must contain the `MCP Raw Data` worksheet. This worksheet writes the SellerSprite MCP raw return data or compact field expansion results used by this report for verification. Before writing, sensitive information such as API keys, tokens, real secret-key URLs, and local absolute paths must be removed.
### 2. Category Product Selection Analysis
```text
# Analyze US station phone stand category
/category-select "phone stand" US
# Analyze UK station laptop backpack category
/category-select "laptop backpack" UK
```
Report saved:
```text
category-reports/{category name}_{SITE}_{YYYYMMDD}/
├── report.md
├── data.json
└── dashboard.html
```
Five-dimensional scoring model:
| Dimension | Score | Scoring logic |
|---|---:|---|
| Market size | 20 | Sales, sales revenue, demand depth |
| Growth potential | 25 | New product opportunities, trends, demand expansion |
| Competition intensity | 20 | Brand concentration, review barriers, price war |
| Entry barriers | 20 | Supply chain, certification, patents, differentiation difficulty |
| Profit space | 15 | Average price, CPC, return risk, cost space |
Comprehensive rating:
| Total score | Rating | Suggestion |
|---:|---|---|
| 80-100 | Excellent | Strongly recommended to enter |
| 70-79 | Good | Can consider entering |
| 50-69 | General | Enter with caution |
| 0-49 | Poor | Do not recommend entering |
### 3. Keyword In-Depth Research
```text
/keyword-research {ASIN} US
```
Report saved:
```text
keyword-reports/{ASIN}_{SITE}_{YYYYMMDD}/
├── report.md
├── keywords.csv
├── negative_words.txt
├── keywords_*.csv
└── categorized_summary.json
```
8-dimensional intelligent classification:
| Dimension | Description | Application strategy |
|---|---|---|
| NEGATIVE | Negative/sensitive words | Direct negation |
| BRAND | Brand words | Competitor play or negation |
| MATERIAL | Material words | Precise matching |
| SCENARIO | Scenario words | Split ad groups by scenario |
| ATTRIBUTE | Attribute modifier words | Long-tail precise matching |
| FUNCTION | Functional words | Broad matching expansion |
| CORE | Core product words | Main word placement |
| OTHER | Others | Supplementary word embedding or manual review |
### 4. Comment In-Depth Analysis
```text
/review-analysis {ASIN} US
```
Report saved:
```text
review-analysis-reports/{ASIN}_{SITE}_{YYYYMMDD}/
├── report.md
└── data/
├── product_detail.clean.json
├── reviews_positive.clean.json
├── reviews_negative.clean.json
└── negative_reviews_analysis.json
```
6-dimensional pain point analysis framework:
| Dimension | Identified content | Severity |
|---|---|---|
| Product defects | Damage, durability, failure | High/medium/low |
| Structure/assembly issues | Loose parts, interface fracture, installation difficulties | High/medium/low |
| Design/functional defects | Complex operation, missing functions, poor experience | High/medium/low |
| Appearance/material issues | Odor, scratches, color difference, material discomfort | High/medium/low |
| Description inconsistency | Size, color, functional expectation deviation | High/medium/low |
| Service/logistics issues | Second-hand/refurbished goods, missing accessories, return difficulties | High/medium/low |
### 5. Product Selection In-Depth Research
```text
# Analyze a product direction
/product-research "magnetic phone tripod" US
# Analyze UK station product direction
/product-research "laptop backpack" UK
```
Report saved:
```text
product-research-reports/{product keyword}_{SITE}_{YYYYMMDD}/
├── report.md
├── data.json
└── dashboard.html
```
Complete process:
```text
Information collection → Data collection → Attribute labeling → Cross-analysis → Competitor and VOC → Barrier assessment → Product selection decision → Report output
```
### 6. Product Planning
```text
# Output product planning based on product direction
/product-planning "camera strap" US
# Organize planning table based on existing research results
/product-planning "magnetic phone tripod" US
```
Report saved:
```text
product-planning-reports/{product planning}_{SITE}_{YYYYMMDD}/
├── report.md
├── planning.xlsx
├── assumptions.md
└── raw/
```
Planning process:
```text
Intended product → Market analysis → Competitor analysis and optimization strategy → SWOT → ABA/keywords → MCP raw data → Data source → Enter/observe/abandon decision
```
## Supported Sites
| Site | Code | Site | Code |
|---|---|---|---|
| US | US | Japan | JP |
| UK | UK | Spain | ES |
| Germany | DE | Italy | IT |
| France | FR | Canada | CA |
| India | IN | Mexico | MX |
| Australia | AU | UAE | AE |
| Brazil | BR | | |
Note: SellerSprite MCP tool schema uses `UK` to represent the UK station. If users input the old habit `GB`, execute skill should standardize to `UK` and then call MCP.
## SellerSprite MCP Interface
### Product Related
| Interface | Function |
|---|---|
| `asin_detail` | Product details |
| `asin_sales_trend` | Sales/sales revenue/price trend |
| `asin_prediction` / `bsr_prediction` | Sales or ranking prediction |
| `keepa_info` | Historical price/ranking auxiliary information |
| `asin_coupon_trend` / `asin_detail_with_coupon_trend` | Coupon/promotion trend |
| `review` | User reviews |
| `competitor_lookup` | Competitor search and comparison |
| `product_research` | Product search and selection research |
### Traffic and Keyword Related
| Interface | Function |
|---|---|
| `traffic_keyword_stat` | ASIN traffic word overview statistics |
| `traffic_keyword` | ASIN reverse query traffic words |
| `traffic_source` | Natural/recommended/ad traffic structure |
| `traffic_extend` | Competitor/keyword traffic extension |
| `traffic_listing` | Traffic or keyword-related Listing |
| `keyword_miner` | Keyword demand, competition, PPC, purchase rate, etc. |
| `keyword_research` | Keyword market screening |
| `keyword_research_trends` | Keyword trends |
| `keyword_order` | ABA ranking/reverse query |
| `aba_research_trend` / `aba_research_monthly` / `aba_research_weekly` | ABA trends and market patterns |
### Product Category Related
| Interface | Function |
|---|---|
| `product_node` | Get nodeId by category name search |
| `market_research_statistics` | Market statistics by category |
| `market_research` | Market screening by category |
| `market_product_demand_trend` | Demand trend by category |
| `market_price_distribution` | Price band distribution |
| `market_rating_distribution` | Rating distribution |
| `market_product_concentration` | Product concentration |
| `market_brand_concentration` | Brand concentration |
| `market_seller_concentration` | Seller concentration |
| `market_seller_country_distribution` | Seller country distribution |
| `market_seller_type_concentration` | Seller/FBA/FBM/self-operated structure |
| `market_listing_date_distribution` / `market_listing_trend_distribution` | Listing time and new product trend |
| `market_ratings_count_distribution` | Comment count barrier distribution |
| `market_ebc_distribution` | EBC/A+ coverage |
### Trademark/IP and Trend Supplement
| Interface | Function |
|---|---|
| `trademark_country_list` / `trademark_list` / `trademark_stats` / `trademark_detail` | Trademark search and risk auxiliary judgment |
| `google_trend` | Google search trend auxiliary verification |
See [references/sellersprite-mcp-api.md](references/sellersprite-mcp-api.md) for details.
English documentation is available at [docs/sellersprite-mcp-usage.md](docs/sellersprite-mcp-usage.md).
`product-planning` MCP call flow is available at [references/product-planning-mcp-flow.md](references/product-planning-mcp-flow.md).
---
## Data Verification
`amazon-analyse` report must include:
```text
Gemini Eval Harness Data Verification
```
Verification dimensions:
| Verification Item | Description |
|---|---|
| Data Consistency | Price, rating, comment count, sales, category, keyword consistency with original data |
| Aperture Transparency | Source, gap, conflict, and adopted aperture explanation |
| Inference Reliability | Over-inference check |
| Risk Integrity | Product, advertising, comment, and fulfillment risk coverage |
| Suggestion Executability | P0/P1/P2 action basis and verification metrics |
If no model key is available, request Gemini/GLM key from the user. If the user refuses or the network is unavailable, indicate `Failed to execute real dual-model call` and only output manual/rule verification results. Do not describe the results as Gemini/GLM real verification.
---
## Project Structure
```text
amazon-sellersprite-research-MCP-skill/
├── .mcp.json.example
├── .env.example
├── .claude/
│ └── skills/
│ ├── amazon-analyse/
│ ├── category-selection/
│ ├── keyword-research/
│ ├── review-analysis/
│ ├── product-research/
│ └── product-planning/
├── skills/
│ ├── amazon-analyse/
│ ├── category-selection/
│ ├── keyword-research/
│ ├── review-analysis/
│ ├── product-research/
│ └── product-planning/
├── references/
├── templates/
├── scripts/
│ ├── python/
├── reports/
├── category-reports/
├── keyword-reports/
├── review-analysis-reports/
├── product-research-reports/
├── product-planning-reports/
└── README.md
```
---
## Skill Development Principles
1. **YAML frontmatter**: `description` must clarify the trigger scenario.
2. **Progressive disclosure**: `SKILL.md` remains concise, with details in `references/`.
3. **Scripts**: For deterministic execution and reusable product generation.
4. **Templates**: Markdown/Excel output templates in `templates/`.
5. **Security first**: Real keys, raw data, and real reports do not enter the open-source repository; locally generated private Excel can include desensitized `MCP Raw Data` worksheets for business review.
---
## Python Tool Entry
```bash
python scripts/python/sellersprite_skill_tools.py check-package
python scripts/python/sellersprite_skill_tools.py scan-sensitive
python scripts/python/sellersprite_skill_tools.py test-open-source-skills
python scripts/python/sellersprite_skill_tools.py publish-github-api --repo-name amazon-sellersprite-research-MCP-skill
```
---
## FAQs
### Q: ASIN query not found?
A: Use `asin_detail` to verify ASIN existence. If the result is empty:
1. Confirm ASIN format correctness.
2. Confirm site correctness.
3. Use `product_research` to search keywords for candidate products.
### Q: Report garbled characters?
A: Check data parsing and file encoding; all Markdown should use UTF-8.
### Q: Excel report cannot open?
A: Prioritize stable templates and in-cell visualization. Be cautious when modifying `.xlsx` chart XML; verify Excel can open.
### Q: Can generate reports be submitted to GitHub?
A: Not recommended. Real reports, raw JSON, comment data, and private conclusions should not be open-sourced.
### Q: Excel or Markdown primary report?
A: Markdown is the primary report and must be retained. Excel is an optional meeting version add-on output; if outputting Excel, include the `MCP Raw Data` worksheet for MCP data basis.
---
## Version
Current version: `V1.2.0`
Current version uses Python toolchain as default execution entry, retaining SellerSprite MCP six research skills, `product-planning` V1 standard meeting version template, desensitization check, and Gemini/GLM cross-verification specification.
---
## License
MIT License
---
## Acknowledgements
- Thanks to [liangdabiao/amazon-sorftime-research-MCP-skill](https://github.com/liangdabiao/amazon-sorftime-research-MCP-skill/tree/main) for README structure, skill organization, and report directory design reference.
MCP Config
Below is the configuration for this MCP Server. You can copy it directly to Cursor or other MCP clients.
mcp.json
Connection Info
You Might Also Like
No recommendations available