> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cassidyai.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Understand prompt engineering

> Learn how clear context, instructions, examples, and output requirements help Cassidy produce more useful and consistent AI responses.

Write clearer instructions so your Agents and Workflows produce useful, consistent results.

[Prompt engineering explained](https://www.loom.com/embed/c1747fe9482e4a0da5c477a591106a1c)

## What you'll learn

* Why specific instructions produce more reliable responses
* Where prompts live in Agent instructions, chat messages, and Workflow Actions
* How role, process, output format, constraints, and examples clarify the result
* How to tell Cassidy what to do when important information is missing

## Put it into practice

**1. Choose where the prompt belongs**

Put consistent behavior, such as role, process, tone, and rules, in Agent instructions. Put task-specific requests and context in chat messages or Workflow inputs.

**2. Describe a good result**

Define the role, process, output format, and constraints. Add one strong example when it explains your expectations more clearly.

**3. Test and refine**

Tell Cassidy to ask a question or flag a gap instead of guessing. Test several realistic requests, then make examples or boundaries more specific when results vary.

## Keep learning

> **[Write effective prompts](https://docs.cassidyai.com/guides/prompt-engineering.md)**
>
> Use a complete framework for clear, reliable prompts.

> **[Build and configure an Agent](https://docs.cassidyai.com/agents/build-and-configure.md)**
>
> Apply prompt techniques to Agent instructions.

> **[Build a Workflow](https://docs.cassidyai.com/workflows/building.md)**
>
> Use prompts with variables and repeatable Actions.

> **[Build your first Skill](https://docs.cassidyai.com/university/build-your-first-skill.md)**
>
> Turn clear instructions into a reusable playbook.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.