Handling tool execution results in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Tool Execution Result Types

When an AI agent executes a tool, it receives a result. Which of the following best describes the typical types of results an agent should expect and handle?

A Only numerical outputs representing scores or probabilities

B Only boolean values indicating success or failure

C Textual outputs, error messages, or structured data like JSON

D Raw binary data without any structure

💡 Hint

Think about the variety of outputs tools can produce, including errors and structured responses.

Agents often receive diverse outputs from tools, including text, error messages, or structured data like JSON. Handling these correctly is essential for robust behavior.

❓ Predict Output

intermediate

Output of Tool Execution Result Parsing

What will be the output of the following Python code that simulates handling a tool execution result?

Agentic AI

tool_result = '{"status": "success", "data": {"value": 42}}'
import json
parsed = json.loads(tool_result)
output = parsed.get('data', {}).get('value', None)
print(output)

A None

B 42

C KeyError

D SyntaxError

💡 Hint

Look at how json.loads parses the string and how get() methods are chained.

The JSON string is parsed into a dictionary. The nested get calls safely access 'value' inside 'data', which is 42.

❓ Model Choice

advanced

Choosing a Model to Handle Tool Execution Errors

You want to build an AI agent that can robustly handle errors from tool executions and decide the next action. Which model architecture is best suited for this task?

A A recurrent neural network (RNN) that processes sequences of tool outputs and errors

B A simple feedforward neural network with fixed input size

C A convolutional neural network (CNN) designed for image data

D A k-nearest neighbors (KNN) model using static feature vectors

💡 Hint

Consider the nature of tool execution results as sequences or logs over time.

RNNs are good at handling sequences, such as logs of tool outputs and errors, enabling the agent to learn patterns and decide next steps.

❓ Hyperparameter

advanced

Hyperparameter to Tune for Tool Result Interpretation Speed

You have an AI agent that interprets tool execution results in real-time. Which hyperparameter adjustment would most directly improve the speed of interpreting these results?

A Add dropout layers to prevent overfitting

B Increase the learning rate during training

C Increase the batch size during training

D Reduce the model's number of layers or parameters

💡 Hint

Think about model complexity and inference speed.

Reducing the number of layers or parameters decreases model complexity, which speeds up inference and result interpretation.

🔧 Debug

expert

Debugging Tool Execution Result Handling Code

What error will the following Python code raise when handling a tool execution result?

Agentic AI

tool_result = None
output = tool_result.get('data', {}).get('value', 0)
print(output)

A AttributeError: 'NoneType' object has no attribute 'get'

B KeyError: 'data'

C TypeError: 'NoneType' object is not subscriptable

D No error, prints 0

💡 Hint

Check what happens when you call get() on None.

Since tool_result is None, calling get() on it raises an AttributeError because None has no get method.