Foundations of LLM-Based Agents

Define an LLM agent and distinguish it from a plain LLM call, a chain, and a workflow.

Describe the core agent loop: perceive, reason, act, observe, repeat.

Explain the ReAct (Reason and Act) prompting pattern for tool use.

Compare agent planning strategies: single-step, plan-and-execute, and reflexion-style self-correction.

Explain how an LLM selects, calls, and interprets results from external functions (tools).

Identify common agent failure modes, such as hallucinated tool calls and infinite loops.

Describe high-level mitigation strategies for common agent failures.