Amazon Sorftime MCP Skills

Core Capabilities#

This project provides five Claude Code slash-command skills for Amazon cross-border e-commerce sellers:

Competitor Analysis (/amazon-analyse)#

• Params: {ASIN} {SITE}
• Dimensions: product details, reviews, traffic keywords, sales/price/rank trends, TikTok videos, 1688 procurement costs
• Output: Markdown, HTML dashboard, Excel, JSON

Category Selection (/category-selection)#

• Params: "{category}" {SITE} [--limit N]
• Data: Category Top100 + real-time statistics
• 5-dimensional scoring model (100 points): market size (20), growth potential (25), competition intensity (20), entry barriers (20), profit margin (15)
• Ratings: 80+ Excellent / 70-79 Good / 50-69 Fair / 0-49 Poor

Keyword Research (/keyword-research)#

• Params: {ASIN} {SITE}
• Scale: 1500+ keywords
• 8-dimensional classification: negative, brand, material, scenario, attribute, function, core, other
• Direct PPC output: negative keyword list, exact match groups, scenario ad groups, broad match groups, plus CSV keyword files

Review Pain-Point Mining (/review-analysis)#

• Params: {ASIN} {SITE}
• 6-dimensional pain-point framework + service risk alerts (e.g., used/defective >5% as danger threshold)
• Dual-track solutions: product improvement suggestions + customer service scripts

LLM-Driven Product Research (/product-research)#

• Params: "{keyword}" {SITE}
• Full pipeline: info collection → data gathering → attribute labeling → cross-analysis → competitor & VOC → evaluation → report output

Cross-Platform Data Coverage#

• Amazon: 14 marketplaces (US/GB/DE/FR/IT/ES/JP/IN/CA/MX/AU/AE/BR/SA)
• TikTok: 6 markets (US/GB/MY/PH/VN/ID) for influencer video analysis
• 1688: China wholesale platform procurement cost analysis

Architecture#

The project connects to Sorftime remote services via MCP (Model Context Protocol) over Streamable HTTP. Sorftime API returns SSE-format data parsed by Python scripts. Each skill uses SKILL.md as entry point (with YAML frontmatter), follows Progressive Disclosure, encapsulates deterministic logic in scripts/, and stores report templates in assets/. Code composition: HTML 52.9% (dashboard templates), Python 47.1%. Includes a built-in skill-creator tool for creating new skill templates.

Sorftime MCP API covers: product details, sales/price/rank trends, reviews (up to 100), traffic keywords, competitor keyword layouts, product search/filter, keyword long-tail expansion, category name search, category real-time reports (Top100+stats), category tree, category historical trends, TikTok product search, TikTok video analysis, 1688 cost analysis.

Setup#

• Runtime: Claude Code CLI + Bash shell
• Python dependency: pip install xlsxwriter for some skills
• Single config: .mcp.json with Sorftime API Key
• Reports output to dedicated directories per skill

Open Questions#

• Sorftime API Key acquisition process and pricing not publicly detailed
• Sorftime MCP service stability and SLA not committed
• Minimum Claude Code CLI version unspecified
• Python version compatibility not declared
• v2.5/v2.6 dates marked as 2026, possibly typos or pre-release
• Single contributor (liangdabiao, 13 commits), long-term maintenance sustainability uncertain