Run Your First Real-World Task

The best way to understand Craftly Robot is to give it a real objective.

You do not need to tell Robot every internal step.

Instead, describe:

What you want + what matters + what constraints apply.

Structure of a Good Request

A useful request usually contains:

• Goal: What are you trying to accomplish?
• Requirements: What must be true for the result to be useful?
• Constraints: What should Robot respect or avoid?
• Schedule: When does the work need to happen?
• Budget: What financial limit or expectation applies?
• Preferences: Are there additional conditions that matter to you?

Example

Less useful:

“Find me a teacher.”

More useful:

“Find a BUET teacher for a Class 9 student who can teach four days a week for a monthly salary of ৳5,000.”

The second request gives Robot enough context to determine what kind of discovery and coordination is required.

What Happens After You Ask?

A typical workflow can look like:

The exact workflow depends on the task.

A simple request may only require information retrieval.

A more complex request may require multiple agents, multiple rounds of communication, and a human decision before any consequential action can occur.

When Robot Needs Clarification

A good agent should not blindly act on an incomplete instruction.

For example:

“Find someone to help me with my studies.”

Robot may need to ask:

• What subject?
• What level?
• What schedule?
• Online or in person?
• What budget?
• What qualifications matter?

This is part of the agentic workflow.

Clarification is not failure. It is how the system avoids making unnecessary assumptions.

Understanding the Result

When Robot returns a result, you should be able to understand:

• What it found
• Why the result may be relevant
• Which requirements were satisfied
• Which requirements remain uncertain
• What action would happen next
• Whether your approval is required

This makes the transition from AI output → real-world action understandable and reviewable.