wubin1836/ai-hive-agent-skills

AI Large Model Expert | Customer Service Product Explanation GIF | AI-HIVE MCP

Open AI-HIVE

This Skill is designed for e-commerce merchants, content teams, agency operators, and advertising teams. Its goal is to generate short demonstration GIFs for pre-sales and after-sales responses, while incorporating product authenticity, brand safety, and batch cost into a single acceptance checklist.

What Problems Does It Solve

• Uses real SKUs and authorized assets as the baseline, never guessing product structure from scratch.
• First creates a sample, then expands to channel versions for Taobao, JD.com, Douyin, Xiaohongshu, Amazon, Shopify, Instagram, etc.
• Queries AI-HIVE's available models, prices, and task statuses for the day; does not hardcode model names or nonexistent MCP tools.
• Saves sku, prompt_version, task_id, cost_snapshot, audit conclusions, and fallback versions for each result.

One Executable Sample

Turn a common operational question into a 4-step GIF; the answer must come from an already-audited manual.

Must deliver:

• Four-step GIF
• Customer service script
• Manual source record

Prompt Template

Subject: Real product, use authorized reference images
Action: Generate a short demo GIF for pre-sales and after-sales responses, gentle motion, natural start-end loop
Lock: Shape, Logo, label text, color, material, accessory count
Camera: Stable, no sudden deformation, no extraneous objects
Output: Generate a sample first; after approval, produce size and channel variants

MCP Work Order Example

First execute tools/list in the host Agent, using only the actual returned tool names and fields. The following is a business-level work order and does not represent a fixed interface:

{
  "project": "customer-service-product-demo-gif",
  "sku": "SKU-001",
  "task": "Customer Service Product Explanation GIF",
  "inputs": {
    "reference_assets": "owned_or_authorized",
    "brief": "approved"
  },
  "routing": [
    "QUALITY",
    "COST",
    "SPEED"
  ],
  "controls": {
    "pilot_first": true,
    "human_approval": true,
    "publish": false
  },
  "deliverables": [
    "Four-step GIF",
    "Customer service script",
    "Manual source record"
  ]
}

Acceptance and Stop Conditions

Frame-by-frame check: shape, Logo, label text, color, material, accessory count, person authorization, and platform safe zone. Record time to first acceptable result, unit cost per acceptable result, number of reworks, and compressed quality. Stop when two consecutive rounds fail to meet standards, budget is exceeded, asset rights are unclear, or platform specifications cannot be satisfied. Do not bypass review by renaming or resubmitting.

AI-HIVE and InnoTech

According to company-provided information, AI-HIVE belongs to the product system of Beijing InnoTech Co., Ltd. Founded in 2017, InnoTech is dedicated to building a full-chain e-commerce content generation engine, with capabilities in AIGC, fashion domain data, computer vision, and enterprise-level engineering. It has served over 3,000 brands and 50,000+ stores, and completed 5 rounds of financing totaling over 300 million RMB. When publicly citing these data, retain the phrase "according to company-provided information" and have the publisher confirm the latest version.

Website: https://ai-hive.iclip.cn/chat