What is the agentic loop in Claude code?	A cycle where Claude code gathers context, interacts with the model to get text or tool calls, takes action (like editing files or running commands), verifies results, and repeats until the prompt is satisfied.
How does Claude code determine if a task is complete?	It verifies the results of its actions and checks if they achieve what the original prompt set out to do; if not, it runs the loop again.
What happens when Claude code reaches its context window limit?	It automatically compacts the conversation by determining what can be removed from the context window and what can be summarized to bring it back down.
What role do tools play in how Claude code works?	Tools are the backbone of agents; they allow Claude code to determine when to execute code (like reading files or searching the web) to get closer to completing a task.
What are the three permission modes available in Claude code?	Default (asks permission before editing files or running commands), Auto (accepts file edits without asking but still asks for commands), and Plan (uses read-only tools to compile an action plan first).
How does Claude code differ from typical AI chat applications?	Claude code can take action by executing tools and commands, verify its own work, and iterate until tasks are complete, rather than just inputting and outputting text.
What is semantic searching used for in Claude code?	Claude code uses semantic searching to determine when to call a tool and retrieve its output.
Why should you be cautious when skipping permission modes?	Giving Claude code free reign to run commands means mistakes could be harder to catch before they happen.
What can Claude code access in your terminal?	It can read your code base, take action on files and commands, and verify its own work.
How can you guide Claude code during the agentic loop?	You can add context, interrupt the loop, or steer the model to help guide it towards your end goal.