TikTok Shop product research tool for Codex and WorkBuddy

The xplaai TikTok Shop product research tool is a packaged Skill for Codex and WorkBuddy. It turns a market, category and keyword brief into a reviewable candidate list, product evidence, selling-point hypotheses and content angles that a human operator can approve.

View the public workflow or browse Commerce Skills.

What the Skill does

The workflow helps a team repeat the same research structure rather than collecting random screenshots and links.

The output can include product images, source links, evidence fields, selling-point hypotheses and content angles. A shortlist is a decision aid, not a prediction of sales.

Start with a dry-run

Install the package for Codex or WorkBuddy, then ask the Agent to run its environment check.

Before the first data request, the dry-run should list:

1. selected market and region;
2. category and keywords;
3. page or candidate limits;
4. fields that will be written;
5. the destination board;
6. missing configuration without printing any secret.

Approve the plan only after those values are correct.

Understand what the sales field means

The current TikTok Shop search data can include cumulative sales returned by the data source. It does not provide a strict seven-day date-window parameter.

Therefore:

• do not label cumulative sales as “sales in the last seven days”;
• do not calculate growth from one snapshot;
• store a later snapshot before comparing changes;
• record collection times and the same product identifier;
• treat unavailable fields as unknown instead of inventing values.

The TikTok Shop data API page is the developer route for teams building their own data application.

Use snapshots for tracking

A first snapshot shows the state observed at collection time. Repeated snapshots can support comparisons when the same product, market and fields are tracked consistently.

Changes still need interpretation. A larger cumulative value may reflect elapsed time, reporting differences or product changes. Human review should combine data with product quality, margin assumptions, fulfillment, creative fit and policy risk.

Turn research into a creative brief

After approving a product, convert evidence into:

• one customer problem;
• three supportable selling points;
• common objections;
• required product demonstrations;
• claims that must not be made;
• short-form content angles.

An approved brief can then inform the English Affiliate Video Skill or another commerce workflow. Research and creative generation remain separate approval stages.

Keep the board under your control

The workflow writes to the user’s own Lark or Feishu configuration. Keep credentials local and do not ask the Agent to print them. Public case screenshots demonstrate the structure of an output, not guaranteed growth or a guaranteed winning product.

Frequently asked questions

Does this tool guarantee a winning product?

No. It organizes candidates and evidence for human review. It does not guarantee sales, virality, margin or platform approval.

Is the sales number a seven-day value?

Not by default. Current search results may contain cumulative sales. A time-window trend requires repeated snapshots and a documented calculation.

What inputs should I provide?

Provide the market, category, keywords, filters, review criteria and the output destination configuration.

Where do results go?

The Skill can organize results in the user’s own Lark or Feishu board according to the installed package instructions.

Can the research run repeatedly?

Yes, after configuration and approval. Repeated runs should preserve product identifiers, field definitions and collection timestamps.

Does it support both Codex and WorkBuddy?

Yes. Use the package created for your Agent and follow its installation and doctor instructions.

Build one reviewable shortlist

Open the public case, choose the correct Agent package, run a dry-run and review the first snapshot before scheduling repeated research.