Getting Started: Best Practices

You do not need to memorize every Agent Central tool. Better results come from smaller API keys, reusable workflows, and prompts that tell Claude, ChatGPT, OpenClaw, or any MCP client exactly what answer to produce.

The Five Operator Habits

1Scope API Keys By Workflow

Give each agent or job the smallest useful Agent Central tool surface.

2Share Tightly Scoped Keys, Not Broad Keys

Create separate keys for teammates, contractors, and write-enabled workflows.

3Build A Prompt Library

Turn useful back-and-forth conversations into reusable one-shot prompts.

4Ask For The Shape Of The Answer

Tell the agent the exact table, sorting, caveats, and takeaways you need.

5Be Precise About Dates, Metrics, And Sources

Avoid ambiguous Amazon sales numbers by specifying dates, metrics, grouping, and freshness.

What This Section Teaches

Agent Central exposes a broad seller-data surface across advertising, inventory, catalog, finance, ranking, and fulfillment. Operators get better results by designing repeatable workflows around the business question, not by memorizing tool names.

Start with a scoped key, ask for the exact output shape, and make the prompt explicit about dates, metrics, identifiers, sources, and caveats. The model can handle tool selection more reliably when the workflow is narrow and the final answer is well defined.

Prompt Patterns To Reuse

These examples are intentionally client-agnostic. Use them in Claude, ChatGPT, OpenClaw, or any MCP client connected to Agent Central.

Reverse prompt

We got to the answer I wanted after several back-and-forth steps with Agent Central.

Turn this into a reusable one-shot prompt I can save in my prompt library.

The prompt should:
- Ask for the same final result directly
- Include the right date range behavior
- Specify the output format
- Ask for SKU/product name instead of only ASIN when relevant
- State which data source or caveats should be included
- Be reusable by changing dates, ASINs, SKUs, campaigns, or marketplaces

Ask for the answer shape

Use Agent Central to show last week's top 20 SKUs by ordered revenue.

Return a table sorted by revenue descending.

Include SKU, product title, ASIN, revenue, units, orders, exact date range, data source, and caveats.

After the table, give me the top 3 takeaways and the exact SKUs to inspect.

Date and metric precision

Use Agent Central to show yesterday's ordered sales by SKU and product name. Include total revenue, units, orders, exact date range, data source, and freshness caveats.