ecommerce-competitor-analyzer（电子商务竞争对手分析器）

E-commerce Competitor Analyzer Skill

Quick Start (For AI)

When to use this skill: When user asks to analyze, research, or extract insights from e-commerce products (Amazon, Temu, Shopee).

What you should do:

1. Extract product identifiers (ASINs or URLs) from user input
2. Call the scraper script to get product data
3. Call the AI analysis with the analysis prompt template
4. Output results in BOTH formats: Google Sheets + Markdown

Input examples:

• "Analyze B0C4YT8S6H"
• "Analyze these products: B0C4YT8S6H, B08N5WRQ1Y, B0CLFH7CCV"
• "Research this competitor: https://amazon.com/dp/B0C4YT8S6H"

Output requirements:

• Google Sheets table with: ASIN, Title, Price, Rating, 4 analysis summaries
• Markdown report with detailed 4-dimensional analysis

How AI Should Process Requests

Step 1: Extract Product Identifiers

From user input, extract all ASINs and/or URLs:

Example inputs:

"Analyze these Amazon products:
B0C4YT8S6H
B08N5WRQ1Y
B0CLFH7CCV"

Extract: ['B0C4YT8S6H', 'B08N5WRQ1Y', 'B0CLFH7CCV']

Mixed input handling:

"Analyze B0C4YT8S6H and https://amazon.com/dp/B08N5WRQ1Y"

Extract: ['B0C4YT8S6H', 'B08N5WRQ1Y'] (extract ASIN from URL)

Step 2: Batch Scrape Product Data

For each product identifier:

1. Detect platform (use scripts/detect-platform.js if available)
2. Call appropriate scraper (Amazon: scripts/scrape-amazon.js)
3. Use Olostep API with configured API key from .env

Batch processing pattern:

// Process all products in parallel
const products = ['B0C4YT8S6H', 'B08N5WRQ1Y', 'B0CLFH7CCV'];
const results = await Promise.allSettled(
  products.map(asin => scrapeAmazon(asin))
);

// Handle failures gracefully
const successful = results.filter(r => r.status === 'fulfilled');
const failed = results.filter(r => r.status === 'rejected');

Step 3: Batch AI Analysis

For each successfully scraped product:

1. Read the analysis prompt from prompts/analysis-prompt-base.md
2. Replace product data placeholders in the prompt
3. Call Gemini API (model: gemini-3-flash-preview)
4. Extract structured analysis results

Analysis framework (4 dimensions):

1. 文案构建逻辑与词频分析 (The Brain) - Copywriting strategy & keywords
2. 视觉资产设计思路 (The Face) - Visual design methodology
3. 评论定量与定性分析 (The Voice) - Review sentiment analysis
4. 市场维态与盲区扫描 (The Pulse) - Market positioning & blind spots

Step 4: Generate Dual Format Output

Format 1: Google Sheets (Structured Data)

Write to Google Sheets with columns: | ASIN | 产品标题 | 价格 | 评分 | 文案分析摘要 | 视觉分析摘要 | 评论分析摘要 | 市场分析摘要 |

Sheet selection priority:

1. User explicitly specified Sheet ID/Name/URL
2. Default from .env (GOOGLE_SHEETS_ID)
3. Ask user to provide Sheet ID

Format 2: Markdown Report (Detailed Analysis)

Generate file: 竞品分析-YYYY-MM-DD.md

Structure:

# Amazon Competitor Analysis Report

## Analysis Overview
- Products analyzed: 3
- Analysis date: 2026-01-29
- Total time: ~5 minutes

---

## Product 1: B0C4YT8S6H

### Basic Information
- Title: [Product title]
- Price: [Price]
- Rating: [Rating]

### Copywriting Strategy & Keyword Analysis
[Full analysis...]

### Visual Asset Design Methodology
[Full analysis...]

### Customer Review Analysis
[Full analysis...]

### Market Positioning & Competitive Intelligence
[Full analysis...]

---

File Structure

ecommerce-competitor-analyzer.skill/
├── SKILL.md                                # This file (AI instructions)
├── platforms.yaml                          # Platform configurations (URL patterns, regex)
├── .env.example                            # Configuration template (API keys)
├── prompts/                                # AI prompt templates
│   └── analysis-prompt-base.md            # Base analysis framework (from n8n)
├── scripts/                                # Processing scripts
│   ├── detect-platform.js                 # Platform detection utility
│   ├── scrape-amazon.js                   # Amazon scraper (Olostep API)
│   └── batch-processor.js                 # Batch processing engine
└── references/                             # Documentation
    └── n8n-workflow-analysis.md           # n8n workflow insights

Configuration Files

platforms.yaml

Contains platform-specific configurations:

• URL patterns for platform detection
• ASIN extraction regex patterns
• Scraper API endpoints
• Data extraction patterns

Key sections:

platforms:
  amazon:
    url_patterns: ["amazon.com", "amazon.co.uk", ...]
    asin_regex:
      standard: "/dp/([A-Z0-9]{10})"
    scraper:
      provider: "olostep"
      api_endpoint: "https://api.olostep.com/v2/agent/web-agent"

.env.example

Template for required API keys:

OLOSTEP_API_KEY=your_olostep_api_key_here
GEMINI_API_KEY=your_gemini_api_key_here
GOOGLE_SHEETS_ID=YOUR_GOOGLE_SHEETS_ID_HERE

Critical: Always check if .env file exists and contains required keys before processing.

Analysis Prompt Template

The AI analysis uses a proven 4-dimensional framework. The exact prompt is stored in: prompts/analysis-prompt-base.md

Key sections:

1. Role: 10-year experienced Amazon Operations Director & Brand Strategist
2. Goal: Deep scan of product listing to extract strategic insights
3. Output Structure:

- Part 1: 文案构建逻辑与词频分析 - Part 2: 视觉资产设计思路 - Part 3: 评论定量与定性分析 - Part 4: 市场维态与盲区扫描

Important: Use the prompt EXACTLY as provided in the template without modifications.

API Services

Olostep API (Web Scraping)

• Purpose: Scrape Amazon product pages with rendered JavaScript
• Endpoint: https://api.olostep.com/v2/agent/web-agent
• Cost: 1000 free requests/month, then $0.002/request
• Key param: comments_to_scrape: 100 (matching n8n config)

Google Gemini API (AI Analysis)

• Purpose: Generate comprehensive product analysis
• Model: gemini-3-flash-preview (cost-effective)
• Cost: ~$0.001/product
• Alternative: gemini-2-flash-thinking (for complex analysis)

Google Sheets API (Data Storage)

• Purpose: Export structured results
• Authentication: OAuth2 service account
• Cost: Free tier

Error Handling

Batch Processing with Error Isolation

Critical pattern from n8n workflow:

const items = productIdentifiers;
const results = await Promise.allSettled(
  items.map(async (item, index) => {
    try {
      const data = await scrapeProduct(item);
      const analysis = await analyzeWithAI(data);
      return { success: true, index, data: analysis };
    } catch (error) {
      // Single failure doesn't stop batch
      return { success: false, index, error: error.message };
    }
  })
);

// Report results
const successful = results.filter(r => r.status === 'fulfilled' && r.value.success);
const failed = results.filter(r => r.status === 'rejected' || !r.value.success);

console.log(`Processed: ${successful.length} succeeded, ${failed.length} failed`);

Common Errors & Solutions

Platform Detection Logic

function detectPlatform(urlOrId) {
  // Direct ASIN
  if (/^[A-Z0-9]{10}$/.test(urlOrId)) {
    return { platform: 'amazon', id: urlOrId };
  }

  // Amazon URL patterns
  if (/amazon\.(com|co\.uk|de|es|fr|it|ca|co\.jp)/i.test(urlOrId)) {
    const asinMatch = urlOrId.match(/\/dp\/([A-Z0-9]{10})/i);
    if (asinMatch) {
      return { platform: 'amazon', id: asinMatch[1] };
    }
  }

  // Other platforms (future)
  // if (/temu\.com/i.test(urlOrId)) return { platform: 'temu', id: extractId(urlOrId) };

  return null;
}

Implementation Notes

Current Version: Phase 1 MVP

Supported Platforms: Amazon (US only) Input Method: Dialog-based (ASINs or URLs) Output Format: Google Sheets table + Markdown report

Roadmap

• ✅ Phase 1: Amazon MVP (current)
• 🔄 Phase 2: Add Temu & Shopee platforms
• 🔄 Phase 3: Cross-platform comparison
• 🔄 Phase 4: Historical tracking & price alerts

Design Philosophy

This skill follows the error isolation pattern from the n8n workflow:

• Single product failure NEVER stops the entire batch
• Always report both successes and failures
• Provide detailed error messages for debugging

Performance Benchmarks

References

• n8n Workflow: Based on v81 workflow logic
• Platform Config: See platforms.yaml for URL patterns and extraction rules
• Analysis Prompt: See prompts/analysis-prompt-base.md for exact prompt template

Important Reminders for AI

1. ALWAYS extract ALL product identifiers from user input before processing
2. ALWAYS use batch processing with Promise.allSettled for error isolation
3. ALWAYS generate BOTH output formats: Google Sheets + Markdown
4. NEVER modify the analysis prompt - use it exactly as provided
5. ALWAYS validate.env exists before starting processing
6. ALWAYS report processing summary: X succeeded, Y failed
7. If Google Sheets ID is missing, ask user to provide it
8. Use the exact prompt from prompts/analysis-prompt-base.md without any modifications