AgentSkillsCN

product-enrichment

Autonord 电商的完整产品丰富化流程。统筹 RAG 研究、TwoPhaseQA 事实提取、AI 内容生成、TAYA 校验、图片搜索,以及 Shopify 更新。适用于需要 AI 生成内容的产品时使用。

SKILL.md
--- frontmatter
name: product-enrichment
description: Full product enrichment pipeline for Autonord e-commerce. Orchestrates RAG research, TwoPhaseQA fact extraction, AI content generation, TAYA validation, image search, and Shopify update. Use when a product needs AI-generated content.

Product Enrichment Skill

The core skill of the Autonord platform. Transforms a bare product listing into a rich, TAYA-compliant product page.

Pipeline Steps

  1. UniversalRAG — Search verified info from whitelisted sources (manufacturer sites, trusted retailers)
  2. RagAdapter — Transform RAG output into TwoPhaseQA input format
  3. TwoPhaseQA — Extract atomic facts (Simple QA) and generate reasoning (Complex QA)
  4. AI Enrichment V3 — Generate full product content (description, pros, cons, FAQs, accessories)
  5. TAYA Police — Validate content against philosophy rules, fix violations
  6. ImageAgent V4 — Find high-quality product image from trusted sources
  7. Shopify Update — Save HTML + Metafields + Image with ALT text

Payload Format

typescript
{
  productId: string;
  title: string;
  vendor: string;
  productType: string;
  sku: string | null;
  barcode: string | null;
  tags: string[];
  price?: string;
}

Output Metrics

  • RAG sources queried and success rate
  • QA verified facts count
  • V3 confidence score
  • Image found (source and method)
  • TAYA violations fixed
  • Pros/cons/FAQs counts

Triggers

  • webhook: Shopify product create/update
  • cron: Batch processing of unenriched products
  • manual: Single product enrichment from Admin Dashboard