AI Agents Exercises — Practice Coding

1

Name the loop pattern: observe → think → act → observe

// Core AI agent execution cycle

// Called the loop

2

Identify the stop reason when an LLM wants to call a tool

if response.stop_reason == "":

execute_tool(response)

3

Name the prompting strategy that interleaves reasoning and action steps

# Thought: I need to search for X

# Action: web_search({"query": "X"})

# Observation: ...

# Called prompting

4

Complete a tool definition object with its required fields

tool = {

"": "get_weather",

"description": "Get weather for a city",

"input_schema": { "type": "object", ... }

}

5

Name the type of memory stored in the conversation messages array

# Retained within the current session only

# Called memory

6

Identify the architecture where one agent delegates to specialists

# One agent coordinates multiple specialist agents

# Called pattern

7

Name the mechanism that prevents infinite agent loops

for step in range():

result = agent.step()

if result.done: break

8

Identify what makes a good tool description

// The model reads this to decide when to call the tool

{

name: "search_docs",

: "Search documentation for a given topic. Use when you need to find technical information.",

}