Alpine Systems Academy
1 of 5
BASECAMP · PROMPT ENGINEERING

Prompt Engineering

Operational prompt design for repeatable output quality.
Build and validate reusable templates in this class.

Standalone Class Structure

Chat Guide: Alice Pine (AI co-pilot)

Objective: Build reusable prompt templates that produce consistent quality.

Inputs: One recurring task, target audience context, output format requirements.

Steps: Apply formula -> select patterns -> build template -> test -> refine.

Checkpoint: Template performs across multiple inputs without major rework.

Pass Criteria: 3 reusable templates validated with real examples.

Finished work: Prompt library with examples and revision notes.

5 min
Finished-work review
Verify outputs against pass criteria
10 min
The Formula
The anatomy of every good prompt
25 min
The 7 Patterns
Apply a repeatable pattern set
15 min
Build Templates
Create reusable prompts for your work
5 min
Refinement
Commands for when it's not quite right
The Mindset

Prompting Is Structured Briefing

Prompt quality follows briefing quality: context, task, constraints, and format.

What Good Briefing Looks Like

Be specific about what you want

Give context — who you are, what you've tried

State constraints — what to avoid, preferences

Define the format — how you want the output

Common Mistakes

Too vague: "Make this better"

No context: "Write a marketing email"

Too much at once: 500 words + 10 asks

Accepting mediocre: 70% there isn't done

The Structure

The Formula

Every great prompt has these four pieces.
Sometimes explicit, sometimes implied.

The Anatomy of a Great Prompt

Context
+
Task
+
Constraints
+
Format
The Problem
"Create a 15-minute executive presentation outline"

No context.
No constraints.
AI doesn't know your audience, your goals, or how long you have.
It's going to guess — and probably guess wrong.

The Toolkit

The 7 Prompt Patterns

Click any pattern to see the template and examples.
These are your building blocks.

1

Role Assignment

Tell AI who to be.
Different roles produce different outputs.

2

Specific Ask

Be precise.
Vague asks get vague answers.

3

Example Provider

Show, don't tell.
Examples are worth 1000 words.

4

Step-by-Step

Break complex tasks down.
Prevents jumping to conclusions.

5

Constraint Box

Define the boundaries.
AI will fill any space you give it.

6

Output Format

Specify exactly how you want results structured.

7

Iteration Setup

Build in feedback loops.
Make AI ask before it acts.

Pattern Title

Template goes here

Examples

    Hands On

    Build Your Templates

    Select one recurring task.
    Build one template that can be reused without rework.

    Role
    Context
    Task
    Format
    Constraints
    Your assembled prompt will appear here... Ready to copy and paste into ChatGPT, Claude, or any AI.
    Save Your Work

    Template Library

    Store validated prompts in a searchable template library.

    Why This Matters

    The best prompts are the ones you don't have to write twice.
    Every time something works well, save it here.
    Over time, you'll build a library of proven templates that get consistent results.

    Systems Track

    A template you've validated twice is a candidate for something bigger — a Custom GPT that applies it automatically, or an automation that runs it without you pasting anything.
    That's the next rung up.

    When It's Not Right

    Refinement Commands

    Use targeted refinement commands to correct output defects.

    Too Long

    Cut the fluff

    "Shorter. Cut by half." "Just the key points, 3 bullets max."

    Too Short

    Need more detail

    "Expand on [specific point]." "Give me the full version."

    Wrong Tone

    Adjust the voice

    "More casual, like texting a friend." "More formal, this is for executives."

    Wrong Direction

    Redirect entirely

    "Focus more on [X] and less on [Y]." "Start over with this approach: [new]"

    Almost There

    Small fixes

    "Good, but change [specific thing]." "Keep points 1 and 3, redo point 2."

    Need Options

    Get alternatives

    "Give me 3 different approaches." "What would a [CFO/skeptic] say?"
    First Step to Persistence

    Use Your Settings

    Use platform settings to reduce repeated setup in new conversations.

    Default Chat Limitation

    Most platforms now carry some memory between chats automatically.
    But ambient memory is inconsistent — it might have the right context, it might not, and you can't always tell which until the output is off.
    Custom Instructions and Settings give you a baseline you set once and control directly: your role, your tone, your standing preferences.
    No guessing whether the AI actually remembered.

    Operator move: Use the built-in settings to give AI a baseline understanding of you.

    ChatGPT: Custom Instructions

    Settings → Personalization → Custom Instructions

    "What would you like ChatGPT to know about you?" — Your role, industry, what you work on

    "How would you like ChatGPT to respond?" — Tone, format preferences, things to avoid

    Claude: Settings

    Your Profile → Settings

    Tell Claude about yourself, your work, and how you prefer responses.
    It applies to all new conversations.

    Try It About You:
    Try It How You Want Responses:
    Before the next class

    Checkpoint Gate

    Complete the checklist and log finished work in the prompt library.

    Pass Criteria

    1) Three templates documented for recurring tasks

    2) At least two test runs per template

    3) Quality improves after refinement commands

    4) Template library saved for reuse

    Build 3 reusable templates

    For tasks you do repeatedly.
    Test them.
    Refine until consistent.

    Use at least 4 of the 7 patterns

    Role, Specific Ask, Examples, Step-by-Step, Constraints, Format, Iteration.

    Do 3 refinement cycles on one output

    Don't accept 70%.
    Push until it's exactly what you need.

    Save your best prompts

    Start a doc, note, or folder.
    You're building a library.

    Set up your Custom Instructions

    ChatGPT Settings or Claude Profile.
    Fill in both boxes.
    Use it for a week.

    The habit to build:

    "What pattern would work here?"
    Wrap-Up

    Your Prompting System Is Built

    You now have a reusable framework for turning vague asks into repeatable prompts, examples, and revision loops.

    Formula
    Context, task, constraints, format
    Patterns
    Role, examples, constraints, iteration
    Templates
    Reusable prompts for recurring tasks
    Refinement
    Revise until the output is deployable

    Before you close this class, export your notes and save at least one prompt template you can use again this week.

    Alpine Systems
    Certificate of Completion

    Basecamp Foundations

    Class 2: Prompt Engineering

    You can now write structured prompts, apply repeatable prompt patterns, and build reusable templates for real work.

    What you can do now
    Prompt Formula Pattern Library Template Design
    Explore the Full Program
    academy.thealpinesystem.com