BYU-Idaho AI Prompt Library

A guide for BYU-Idaho employees and students. The prompt library is available for all faculty, employees, and students at BYU-Idaho at promptlibrary.byui.edu.

What is the AI Prompt Library?

The AI Prompt Library is a web application built for the BYU-Idaho community. It provides a browsable, searchable catalog of AI resources designed to help students, faculty, and staff get better results from AI tools like ChatGPT, Copilot, Gemini, and Claude.

The library houses three types of resources:

• Prompts — Ready-to-use text templates for AI tools. A well-crafted prompt helps you get more consistent, higher-quality responses from an AI assistant. Prompts in the library include placeholder variables in [brackets] that you customize before use. For example, a prompt might include [course name] or [assignment topic] — you replace those placeholders with your own details before sending the prompt to an AI tool.
• Agent Skills — Reusable instruction sets that define specialized capabilities for AI agents, following the AgentSkills.io specification. Think of these as building blocks for more advanced AI workflows. If you work with AI agents or plan to, skills give you a head start with tested, structured instructions.
• Agent Configurations — Complete agent definitions that describe how an AI agent should behave, what tools it can use, and what tasks it handles. These follow the AGENTS.md specification and provide a full blueprint for setting up an AI agent.

AI tools are increasingly part of academic and administrative work across campus. When everyone starts from scratch writing prompts, the results are inconsistent and the learning curve is steep. The Prompt Library solves this by giving the BYU-Idaho community a shared space to publish what works and learn from each other.

Instead of figuring out the best way to ask an AI tool to help with lesson planning, research summaries, data analysis, or administrative tasks, you can search the library for a proven prompt, customize it, and use it immediately. Contributors share their expertise, and the entire campus benefits.

How content is organized

Every resource in the library is tagged with metadata that makes it easy to find what you need:

• Category — Academic, research, writing, coding, administrative, creative, or data analysis
• Audience — Student, faculty, staff, or general
• AI Tool — ChatGPT, Copilot, Gemini, Claude, or generic (works with any tool)
• Difficulty — Beginner, intermediate, or advanced

All content is community-contributed. Anyone with a BYU-Idaho account can sign in and start creating resources to share with the campus.

Understanding the Three Resource Types

The Prompt Library organizes AI resources into three types, each serving a different level of complexity. Think of them as a progression: prompts are individual questions, skills are repeatable procedures, and agents are complete AI assistants with defined expertise.

Prompts

A prompt is a piece of text you give to an AI tool to get a specific result. The quality of what you get back depends heavily on how you ask. A vague question produces a vague answer. A well-structured prompt with clear instructions, context, and constraints produces a focused, useful response.

Prompts in the library are ready-to-use templates. They include placeholder variables in [brackets] that you replace with your own details before submitting. For example:

Summarize the key concepts from [topic] for a [course level] course. Present the summary as a bulleted list with no more than [number] items. Use language appropriate for [audience].

You would replace [topic], [course level], [number], and [audience] with your specifics, then send the completed prompt to your preferred AI tool.

Each prompt in the library also includes:

• Description -- A short summary of what the prompt does and when to use it
• Usage Tips -- Explains why the prompt is effective and how to get the best results
• Example Output -- A sample of what the AI tool produces when given the prompt
• Metadata -- Category, audience, AI tool, and difficulty level to help you find the right prompt quickly

Prompts are the simplest resource type and the best starting point for anyone new to working with AI tools.

Skills (SKILL.md)

A skill is a structured set of instructions that teaches an AI tool how to perform a specific, repeatable task. While a prompt is a single question, a skill is more like a standard operating procedure -- it walks the AI through multiple steps, tells it what information to gather, and defines exactly how to format the result.

Skills follow the AgentSkills.io specification, an open standard for defining reusable AI capabilities. This means skills created in the Prompt Library are portable and can be used with any AI platform that supports the specification.

A typical skill includes:

• Skill Name -- A unique, machine-readable identifier (e.g., meeting-notes-summarizer). This is how AI systems reference the skill programmatically.
• Description -- What the skill does and when to use it
• Instructions -- The core of the skill, written in Markdown. This contains the step-by-step logic the AI follows, including what questions to ask the user, how to process the input, what format to use for the output, and what rules to follow.
• Scripts -- Optional code files that support the skill (e.g., a Python script for data processing)
• References -- Optional documentation or reference material the skill depends on
• Assets -- Optional files such as templates or images

Here is what a simplified skill's instructions might look like:

# Meeting Notes Summarizer

## Purpose
Convert unstructured meeting notes into actionable summaries.

## Instructions
1. Accept the raw meeting notes from the user
2. Ask for: meeting name, date, and attendees
3. Produce a summary with:
   - Key decisions (numbered, with who made each decision)
   - Action items (table with columns: Task, Owner, Due Date)
   - Unresolved items (bulleted list)
4. Keep the summary under 500 words

## Constraints
- Do not add information that is not in the original notes
- If an action item owner is unclear, flag it as "[Owner unclear -- verify]"

Skills are ideal when you have a task you perform regularly and want the AI to handle it consistently every time -- writing meeting summaries, drafting course syllabi, generating assessment rubrics, or preparing department reports.

Agents (AGENTS.md / CLAUDE.md)

An agent configuration is the most advanced resource type. It defines a complete AI assistant with a specific identity, area of expertise, set of capabilities, and rules for how it interacts with users. Where a prompt asks one question and a skill handles one procedure, an agent is designed for ongoing, multi-turn conversations within a specialized domain.

Agent configurations follow the AGENTS.md specification -- a Markdown-based standard for describing AI agent behavior. A related format, CLAUDE.md, serves the same purpose but is specifically designed for Anthropic's Claude. Both formats use the same principles: define who the agent is, what it can do, what it cannot do, and how it should communicate.

A typical agent configuration includes these sections:

• Identity -- Who the agent is and what it specializes in. This sets the agent's expertise and perspective. For example: "You are a course design assistant for BYU-Idaho faculty. You specialize in backward design methodology."
• Capabilities -- What the agent can help with, listed as specific functions. For example: learning outcome development, assessment design, activity sequencing, workload estimation.
• Constraints -- Hard boundaries the agent must respect. These are critical for responsible AI use. For example: "Do not fabricate BYU-Idaho policies," "Defer to instructor judgment on theological content," "Comply with FERPA when handling student information."
• Interaction Style -- How the agent communicates. For example: "Ask 5-8 clarifying questions before generating content," "Present options rather than single answers," "Use a collaborative, peer-level tone."
• Example Interactions -- Sample conversations that demonstrate the agent in action, showing both user questions and expected agent responses.

Here is a simplified example:

# Academic Advisor Copilot

## Identity
You are an academic advising assistant for BYU-Idaho staff advisors.
You help prepare for student appointments by organizing relevant
information and suggesting talking points.

## Capabilities
- Appointment preparation: organize student context into a briefing
- Communication drafting: help write follow-up emails to students
- Advising notes: structure session notes in a consistent format

## Constraints
- FERPA compliance is non-negotiable -- never share student
  information outside the advising context
- Do not make enrollment decisions -- present options for the
  advisor to evaluate
- Flag when a situation may require referral to another office

## Interaction Style
- Ask what type of appointment before preparing materials
- Present information in scannable bullet points
- When uncertain, say so rather than guessing

Agents are valuable when a department or team needs a specialized AI assistant that follows consistent rules and stays within defined boundaries. Rather than re-explaining context and constraints every time you start a new conversation, an agent configuration captures all of that once and applies it automatically.

How the three types compare

	Prompt	Skill	Agent
What it is	A single question or instruction	A repeatable, multi-step procedure	A complete AI assistant persona
Typical length	50-500 words	500-2,000 words	1,000-3,000 words
Best for	Quick, one-off tasks	Recurring workflows with consistent output	Ongoing conversations within a domain
Example	"Summarize this article in 3 bullet points"	"Transform meeting notes into a structured summary with action items and owners"	"Act as a course design consultant who helps faculty build syllabi using backward design"
Complexity	Beginner-friendly	Intermediate	Advanced

Why Markdown?

Prompt text, skill instructions, and agent configurations in the library are written in Markdown, a lightweight formatting language widely used across the technology industry. If you have ever written a bulleted list in a chat application or used asterisks for bold text, you have already used something close to Markdown.

What Markdown looks like

Markdown uses simple symbols to format text. Here are the basics:

What you type	What it does
# Heading	Creates a section title
**bold text**	Makes text bold
- item	Creates a bullet list
1. item	Creates a numbered list
`code`	Formats a word as code or a technical term
```	Opens or closes a code block (use three backticks on their own line)

These basics cover the vast majority of what you will need when writing prompts and other resources.

Why it matters for AI

AI tools like ChatGPT, Copilot, Claude, and Gemini all understand Markdown formatting. When you structure a prompt with headings, lists, and emphasis, the AI can better parse the organization of your request. A well-structured prompt with clear sections tends to produce more organized and accurate responses than a wall of unformatted text.

Markdown is also the standard format for agent skills (AgentSkills.io) and agent configurations (AGENTS.md). Using Markdown across all three resource types keeps everything consistent and portable. You can copy content between the library and your AI tools without losing formatting or structure.

You do not need to be a Markdown expert to use the Prompt Library. If you can write a heading and a bulleted list, you have everything you need to get started.

Roles and Permissions

Every user who signs in to the Prompt Library is assigned a role that determines what they can do. There are three roles, each building on the one before it.

Contributor

This is the default role for all new users. As a Contributor, you can:

• Browse and search all public content in the library
• Create your own prompts, skills, and agent configurations
• Edit and delete your own content
• Set your own content to Private (visible only to you) or Public (visible to everyone)

Contributors cannot see other users' private content, mark content as Featured, or manage other users.

Admin

Admins have all the capabilities of a Contributor, plus additional authority to manage content and users across the library:

• View, edit, and delete any content, including other users' private content
• Mark content as Featured, which highlights it on the Featured page for the entire community
• Manage users: view the user list, change user roles, and deactivate accounts

Admin roles are assigned by the system Owner.

Owner

The Owner has full control over the system. There is one Owner, identified at the system level by the system administrator. The Owner has all Admin capabilities plus complete system control, including the ability to assign Admin roles.

Permissions summary

Capability	Contributor	Admin	Owner
Browse and search public content	Yes	Yes	Yes
Create own content	Yes	Yes	Yes
Edit/delete own content	Yes	Yes	Yes
Set own content visibility (Private/Public)	Yes	Yes	Yes
View others' private content	No	Yes	Yes
Edit/delete others' content	No	Yes	Yes
Mark content as Featured	No	Yes	Yes
Manage users	No	Yes	Yes
Full system control	No	No	Yes