AI AGENT OPERATIONS GUIDE

Daily Operations (HuaYoung TK CLI + AI Agent)

Let AI Agents collaborate with you on shop and creator operations through the HuaYoung TK CLI

With the rise of AI Agent frameworks such as OpenClaw, Hermes, Codex, and WorkBuddy, the audience for HuaYoung TK has shifted.

In the past, HuaYoung TK served TK sellers; now, it serves not only TK sellers but also a wide range of AI Agents. In other words, HuaYoung TK will act as the underlying platform that provides various Agents with the data and functional services related to shops and creators.

1.

Configure the AI Agent

Handle creator operations in one sentence — generate an authorized Key and let your AI Agent access HuaYoung TK capabilities through the HuaYoung TK CLI
  1. 1.1

    Generate an Authorized Key

    Log in to the HuaYoung TK portal, go to "My Personal Settings" from the homepage or your personal menu, and click to generate a new authorized Key in the "AI Agent Configuration" area.

    The Key represents your current user identity and permissions. Please keep it safe and do not share it on public channels.
    Generate an AI Agent authorized Key
    Generate an AI Agent authorized Key
  2. 1.2

    Automatically install the HuaYoung TK CLI

    Send the following instruction to your AI Agent (such as Codex or WorkBuddy) and the CLI will be installed automatically:
    Text 
    The official website of the HuaYoung TK CLI is: www.npmjs.com/package/@huayoung/huayoungtk-cli. Please install it for me.
    Install the HuaYoung TK CLI in one sentence through your AI Agent
    Install the HuaYoung TK CLI in one sentence through your AI Agent
    The HuaYoung TK CLI is updated very frequently. When a new version is released, you only need to ask your AI Agent to reinstall it:
    Text 
    The official website of the HuaYoung TK CLI is: www.npmjs.com/package/@huayoung/huayoungtk-cli. Please reinstall it for me.
  3. 1.3

    Complete authorization

    After the CLI is installed, provide your authorized Key to the AI Agent:
    Text 
    Complete authorization, key: xxxxxxxxxxxxxxxxx
    Provide the authorized Key to the AI Agent
    Provide the authorized Key to the AI Agent
  4. 1.4

    Let the AI get to work

    After authorization is complete, the AI Agent can invoke all of HuaYoung TK's automation capabilities.
    Upgrade shop operations from "doing it yourself" to "supervising the AI while it works"
    Upgrade shop operations from "doing it yourself" to "supervising the AI while it works"
2.

Complete Conversation Examples

The following real conversation scenarios show how the AI Agent reads, asks, and responds in operational workflows.
You can copy the questions from the examples below and send them to the AI Agent. In addition, Chapter 3 provides a prompt library for common operational scenarios.
  1. 2.1

    Product Operations Diagnosis

    Let the AI diagnose product health from multiple dimensions — GMV, order count, return rate, and review rating — to find problem products and opportunity products.

    Sort products from the US store over the last 30 days by return rate and review rating, and flag problem items

    User
    Sort products from the US store over the last 30 days by return rate and review rating, and flag the problem items.
    AI
    Done. 28 SKUs in total. 3 problem items:
    · SPF50 Sunscreen: return rate 18% (average 6%), 12 one-star reviews, keywords "allergy" "stinging"
    · Lip Glaze #03: average rating 3.1, 5 photo negative reviews, buyers report "severe color difference"
    · Concealer: 840 units in stock, only 23 orders in 30 days, slow-moving
    25 healthy items. Lip Glaze #07 has a 1.2% return rate and a 4.7 rating — ready to scale up.
    User
    Summarize the negative reviews for the sunscreen and tell me whether the issue is with the product or with shipping.
    AI
    Compiled 12 one-star reviews. Keyword breakdown:
    · "allergy/stinging": 8 → product formulation issue
    · "leaking/broken packaging": 3 → logistics issue
    · "no effect": 1 → expectation mismatch
    Suggestion: ask the supplier to refine the formula for the formulation issue, switch packaging for the logistics issue, and do not pull the product from sale for now.
    User
    Compile the negative reviews for Lip Glaze #03 as well — I'll send them all to the supplier together.
    AI
    Compiled the photo negative reviews for Lip Glaze #03 — 5 in total. Keywords: "severe color difference" "doesn't match the photo"
    Saved to the workspace file: Negative-Review-Summary-20260823, including the issue breakdown for both the sunscreen and the lip glaze.
  2. 2.2

    Creator Research and Outreach

    From sourcing creators in the open sea to analyzing follower profiles, evaluating promotion capability, and creating targeted invitations — the entire workflow is automated.

    Find 15 creators in the TH site's Beauty category with over 30-day GMV above 5,000, and show their follower profiles and promoted category distribution

    User
    Find 15 creators in the TH site's Beauty category with 30-day GMV above USD 5,000, and show their follower profiles and promoted category distribution.
    AI
    Found 15. Sorted by GMV descending:
    · @beautybynune: 82K followers, GMV $12,400, 72% female followers aged 18–34, 68% in Beauty category
    · @makeup_by_mint: 56K followers, GMV $9,800, 65% female followers, mix of Beauty and personal care
    · @skincare_with_nam: 113K followers, GMV $7,200, followers skew 25–44, heavy skincare mix
    The TOP 3 all fit our Beauty category — recommend prioritizing them.
    User
    Create a targeted invitation on the TH site for the top 3, choose the lip glaze SKU-001 as the product, and set the targeted commission to 15%.
    AI
    Invitation target: @beautybynune, @makeup_by_mint, @skincare_with_nam. Product: Lip Glaze SKU-001. Targeted commission: 15%.
    Confirm creation?
    User
    Confirmed. Create it.
    AI
    Targeted invitation plan created ✅
    Plan ID: TP-TH-20260823-001
    The creators will receive the invitation notification within 24 hours.
  3. 2.3

    Intelligent Sample Approval

    Let the AI evaluate across three dimensions — category match, historical promotion GMV for our store, and duplicate-sample risk — and provide approve/hold/reject recommendations.

    Approve the 3, hold 2, reject 1. Go ahead and execute.

    User
    Review the pending sample applications for the US store. Give recommendations based on category match, historical promotion GMV for our store, and duplicate-sample risk.
    AI
    6 applications pending:

    ✅ Recommended to approve (3)
    · @glowwithmia: Beauty category match, historical promotion GMV for our store $4,200, no duplicate samples
    · @skin.byme: highly matched category, historical GMV $2,800
    · @beauty.nune: category match, first application

    ⏸️ Recommended to hold (2)
    · @freshdaily.skin: Beauty category but historical GMV 0, no promotion videos in the last 90 days
    · @jenny_beauty: another store has already sampled the same item — high duplicate risk

    ❌ Recommended to reject (1)
    · @home.joy: home category — total mismatch
    User
    Approve the 3, hold 2, reject 1. Go ahead and execute.
    AI
    Executing…
    Approved: 3 (jobIds: j-001, j-002, j-003)
    Held: 2 (jobIds: j-004, j-005)
    Rejected: 1 (jobId: j-006)
    The RPA will finish within about 2 minutes. I will notify you when complete.
    User
    Is it done? Notify me of the result.
    AI
    All actions complete ✅
    · 3 approved: sample orders generated
    · 2 held: marked as on hold
    · 1 rejected: rejected
    All actions have been synced to the HuaYoung backend.
  4. 2.4

    Link Funnel Diagnosis

    Find the bottleneck along the "creator sourcing → sample → content → GMV" chain and locate the conversion-rate issue.

    Following the chain "creator sourcing → sample → content → GMV", find out where the US store is stuck

    User
    Following the chain "creator sourcing → sample → content → GMV", find out where the US store is stuck.
    AI
    US store 30-day funnel:
    · Invited creators: 120 (normal)
    · Accepted invitations: 48 (40% ✅)
    · Sample applications: 32 (67% ✅)
    · Received & produced: 18 (56% ⚠️)
    · Posted videos: 12 (67% ✅)
    · Generated GMV: 8 (67% ✅)

    Conclusion: stuck at "received → produced". 14 creators received the sample but did not publish content — 7 of them are over 3 days overdue.
    User
    List the creators who received the sample but are overdue in producing content. I want to send them in-app messages as a reminder.
    AI
    7 creators overdue:
    · @makeup.sara (received 7 days ago, 3 days overdue)
    · @skincare.maya (received 6 days ago, 2 days overdue)
    · @glow.luna (received 5 days ago, 1 day overdue)
    … (7 in total)
    Do you want me to send batch follow-up messages?
    User
    Send in-app messages to the first 3 asking whether they received the sample and when they can post the video.
    AI
    3 in-app messages sent:
    · @makeup.sara ✅
    · @skincare.maya ✅
    · @glow.luna ✅
    Message: "Hello — your sample was received X days ago. When can you post the promotion video? Please contact us if you have any issues."
  5. 2.5

    Comment Patrol and Alerts

    Automatically patrol one- and two-star negative reviews, prioritizing those with photos or videos, and generate remediation suggestions.

    Patrol the one- and two-star reviews for the US store over the last 7 days, and prioritize those with photos or videos.

    User
    Patrol the one- and two-star reviews for the US store over the last 7 days, and prioritize those with photos or videos.
    AI
    47 reviews in the last 7 days — 8 low-rating ones:

    🔴 High priority (with photo/video)
    · Lip Glaze #03: 2 one-star reviews with photos, buyers show "severe color difference"
    · Sunscreen: 1 one-star review with video, feedback "red and irritated after use"

    🟡 Medium priority (text only)
    · 5 text negative reviews in total. Keywords: "average effect" "slow shipping"

    Suggestion: confirm the color standard with the supplier for Lip Glaze #03, and stop using the sunscreen and contact quality control.
    User
    Download the photo negative reviews for Lip Glaze #03 — I'll send them to the supplier.
    AI
    Compiled the photo negative reviews for Lip Glaze #03, including buyer handles, review time, and screenshot links. Saved to the workspace as: Negative-Reviews-LipGlaze03-20260823.
    User
    Do the same for the sunscreen photo negative reviews — I'll send them all together.
    AI
    Compiled the sunscreen photo negative reviews — 1 in total, feedback "red and irritated after use".
    The two reports have been merged: Negative-Review-Summary-20260823, including the issue breakdown and screenshots for both the lip glaze and the sunscreen.
3.

AI Operations Prompts

Prompt templates for common operational scenarios. Copy and send them to the AI Agent to get professional responses for each scenario.
  1. 3.1

    Product Operations

    Hits, return rate, and GMV trends — let the AI monitor product health

    Show me the top 5 best-selling products in the US store over the last 30 days, sorted by order count descending.

    List all US-store products with a return rate above 10% — give me the product name and return rate.

    Analyze the reviews for Lip Glaze #03 and tell me what issues the one- and two-star negative reviews focus on.

    List US-store SKUs with stock above 500 units but fewer than 20 orders in 30 days — these are slow-moving items.

    Show me the US store's product GMV trend for the last week, and tell me which categories are growing and which are declining.

  2. 3.2

    Creator Discovery

    Sourcing from the open sea, targeted invitations, and capability research — covering the full creator-sourcing workflow

    Find 20 creators in the US Beauty category with 50K–200K followers and 30-day GMV of at least USD 3,000 from the open sea.

    Show me the follower profile for @beautybynune — I want to see gender, age, and region distribution.

    Check @glowwithmia's promotion capability: GPM, average video views, and live-stream GMV.

    Sort the creators in my store list by GMV descending and show me the details and ROI of the TOP 10.

    Create a targeted invitation plan for these 3 creators — product: Lip Glaze SKU-001, targeted commission: 12%.

    Get me the contact info for @makeup.sara — I want to add her on WhatsApp.

  3. 3.3

    Sample Approval

    Intelligent approval, fulfillment tracking, and in-app follow-ups — hand the sample workflow to the AI

    Review the pending sample applications for the US store. Give recommendations based on category match, historical promotion GMV for our store, and duplicate-sample risk.

    Among the samples shipped this week, who has received the sample but not posted a video? Sort by overdue days.

    Send an in-app message to @makeup.sara asking whether she received the sample and when she can post the video.

    Check @glowwithmia's sample history — has she requested the same item from other stores?

  4. 3.4

    Data Review

    Link diagnosis, ROI analysis, weekly reports — let the AI do the data review

    Which store had the highest GMV this week? Give me a one-sentence conclusion.

    Pull this month's creator ROI data. I want the TOP 5 and BOTTOM 5.

    Generate this week's operations weekly report, including GMV, creator ROI, video order count, and sample conversion rate.

    Show me the voice share of the TOP 20 creators. Is there a single-creator dependency risk?

    Aggregate the collab count and GMV by BD, and see who is efficient at sourcing and who has low output.

  5. 3.5

    Content Monitoring

    Video rankings, orders per thousand views, hit-product mining — AI helps you track content performance

    What are the TOP 5 promotion videos by order count this week? Give me the links, view count, and orders per thousand views (ordersPm).

    Check this video's view count and order count, and tell me how high the orders-per-thousand-views is.

    Analyze the common features of the TOP 5 hit videos and tell me what type of content tends to drive orders.

    How was the live-stream GMV ranking this week? Show me the sessions above 100 and the promotion data.

    List the authorized videos sorted by GMV — I want to see which content performs best.

  6. 3.6

    Comment Patrol

    Pulling negative reviews, prioritizing photos, and remediation suggestions — automatic patrol of one- and two-star reviews

    Patrol the one- and two-star reviews for the US store over the last 7 days, and prioritize those with photos or videos.

    Check the review statistics for Lip Glaze #03 over the last 30 days — show me the rating distribution and trend.

    Compile all US-store negative reviews with videos. I want to look at the issue in each one.

  7. 3.7

    Buyer Operations

    VIP identification, tag management, private message outreach — AI helps you run fine-grained buyer operations

    Show me the buyer list for the US store for the last 30 days, sorted by order amount.

    Mark the top 10 buyers by spend as VIP for follow-up.

    Send a thank-you private message to the US-store buyers who recently placed orders, letting them know they can contact us anytime.

  8. 3.8

    Link Diagnosis

    Funnel bottleneck, conversion analysis, sync status — locate the breakpoints in the full creator-operations chain

    Following the chain "creator sourcing → sample → content → GMV", find out where the US store is stuck.

    Show me the conversion rates of recent invitation plans in the US store — calculate both accepted/invited and promoted/accepted.

    When was the last sync for the US store? Is the data stale?

    Show me all in-progress invitation plans in the US store and flag the ones that are about to expire (within 7 days).