Example: naming study

A marketing agency gets requests every week from businesses about to open: a coffee shop, a hair salon, an online shop. The first thing they need is a name with a free domain. This harness does that job: it looks up the competitors in the city, suggests names, checks their domains, replaces the taken ones and delivers the report in the agency’s format.

It is the flow in the video of this section.

Piece	Component	What for
Harness	Deep Agent	Plans and does the work.
Model	DeepSeek	Decides each step. Any model with tools will do.
Tool	Web Search (SearXng)	Searches the internet, with no keys.
MCP server	Find-A-Domain Mcp	Checks whether a domain is free.
Skill	Stored Skill with `naming-study`	The agency’s method.

1. Create the skill

In Configuration → My Skills, click New Skill and fill in:

Name: naming-study
Description: Our method to suggest the name of a new business and check its domains
Instruction:

# Naming study

When you are asked for the name of a new business, follow these steps in this order:

1. Search the internet for three similar businesses in the same city, so you neither repeat nor imitate their names.
2. Suggest five names made of two joined words, easy to say in English and in Spanish, without the name of the city.
3. Check the .com and .net domains of each name, written in lower case, with no spaces. If both are taken, replace that name with another one and check it too, until you have five names with a free domain. If a check fails, write "not checked": never call it free or taken.
4. Deliver the report in this format:
   - **Competitors**: the three businesses you found, one line each.
   - **Names**: a table with the columns Name, .com, .net and Why it works.
   - **Recommendation**: the name with the most free domains and one sentence on why.

Click Create.

2. Build the agent

Create a new project and add Chat Input. With the + of its output, add a Deep Agent; with the + of its Model, a model (in the example, DeepSeek with its credential); and with the + of its Response, a Chat Output.
In the Deep Agent’s System Prompt, write:

You are the business launch assistant of a marketing agency.
Before you start a task, check whether one of your skills explains how to do it and follow it step by step.
Search the internet and check domains with your tools: never make anything up.
Always answer in English.

The prompt does not explain how the study is done: that is in the skill. That way the same agent works for other jobs if you connect more skills to it.

3. Connect the tools

With the + of the Deep Agent’s Tools, add Web Search (SearXng). It needs no configuration.
Open the palette, go to MCPs → Cloud & DevOps and drag Find-A-Domain Mcp onto the canvas. In its configuration, create a credential (this server is open: giving it a name is enough), choose the check_domain tool and click the AI button on Name and Tld so the agent fills them in. Connect its Tool output to the Deep Agent’s Tools.

4. Connect the skill

In the Deep Agent’s advanced options, turn on Enable Skills. The Skills connection appears.
With the + of Skills, add Stored Skill and choose naming-study in its Skill field.

5. Try it

Open the Playground and type:

I want to open a specialty coffee shop in Manchester. Prepare the naming study for me.

The agent takes a minute or two. In the Executions list on the left you see every step: the agent reads the skill, runs several web searches, checks the domains one by one with the MCP server and checks again the names it had to replace. At the end it delivers the three sections of the skill: the competitors it found, the table of names with their domains, and the recommendation.

If a result does not convince you, do not touch the agent: change the skill. For example, add “avoid Spanish words” or “also check the .org”, and the next run applies it.

Ideas to go further

More skills in the same agent: one for slogans, another for a pricing study. The agent opens the one each job needs.
Deliver the report: add an email or storage tool and one more step in the skill: “save the report in the client’s folder”.
Review: with Enable Rubric and the criterion “every name in the table has at least one free domain”, the agent does not deliver until it is met.
Publish it: deploy a version and use it through an API, as an MCP server or from another agent.