Prep Library

Context Engineering

The discipline of assembling the right context for LLMs and agents — prompting patterns, retrieval, memory, tool use, and evaluation.

Prompting

Overview

Prompting is the craft of instructing a language model so it responds reliably — using role framing, few-shot examples, chain-of-thought, and structured output formats. Well-designed prompts turn an unpredictable model into a dependable component.

Prompting patterns

Role, few-shot, chain-of-thought.

Core

Structured output

JSON, schemas, function calling.

Core

Context assembly

Overview

Context assembly is about deciding what information goes into a model's limited context window and how — budgeting tokens, retrieving relevant knowledge (RAG), and persisting memory across turns. It's the difference between a grounded, on-topic answer and a hallucinated one.

Context window budgeting

Spend tokens where they matter.

Core

Retrieval-augmented generation

Ground answers in your data.

Advanced

Memory & history compaction

Persist what matters across turns.

Advanced

Agents

Overview

Agents let a model take actions, not just generate text — calling tools, running reason-act loops toward a goal, and being evaluated end to end. This group covers how to design agents that make real progress without spiraling or stalling.

Tool use & function calling

Let the model act.

Advanced

Agent loops & guardrails

Plan, act, observe, repeat.

Advanced

Context & agent evals

Measure end-to-end quality.

Advanced

Where this content comes from

For full transparency, this library's content is curated and verified from these sources:

Frontier-lab prompting & agent guidesRetrieval-augmented generation literatureOppZen-authored context-engineering playbooks