Anatomy of a Great Prompt

Reviewed & published by Brayan K

Great prompts are built from a handful of parts: a role, a task, context, a format, constraints, and examples. Learn these six and you can assemble a strong prompt for almost anything.

Part of the free Prompt Engineering course at LearnCodingFast — hands-on lessons with worked examples and the output they print, plus practice exercises and a quick quiz.

You won’t need all six every time. The trick is to include the parts that remove the guesswork for your particular request — the more ambiguous the job, the more parts you add.

What You'll Learn in This Lesson

• The six building blocks of a strong prompt
• What 'role' and 'task' each do
• How context and constraints differ
• Why asking for a format saves time
• How a single example sets the style
• A reusable structure you can fill in

1 The six building blocks

Most strong prompts mix some of these ingredients. Use the ones your task needs:

Think of it like stacking layers. Each layer you add removes one more thing the AI would otherwise have to guess.

2 Watch a prompt come together

Same goal — a lunch idea — but the strong version layers in the parts:

No role, no constraints, no format. You’ll get a generic list.

Five parts, almost nothing left to guess.

Try it for real

A prompt is not one lump of text — it is six labelled pieces stuck together, and you can hold each one separately. Build the whole thing here, then delete FORMAT and CONSTRAINTS and read the difference. Nothing is hidden: the second prompt still asks the question, it just stops saying how long the answer should be and what the model may use.

# The six blocks, as code. Each is just a labelled string; the prompt is their sum.
# Comment one out and re-run to see exactly what the model loses.

parts = {
    "ROLE":        "You are a support agent for a UK bicycle shop.",
    "TASK":        "Answer the customer's question.",
    "CONTEXT":     "Policy: repairs are free within 12 months of purchase.",
    "FORMAT":      "Reply in two sentences, no bullet points.",
    "CONSTRAINTS": "Use only the policy text. If it is not covered, say so.",
    "EXAMPLE":     'Q: "Is my puncture covered?" -> A: "Punctures are wear and tear, so they are not covered."',
}

def build(selected):
    return "\n\n".join(f"{k}:\n{parts[k]}" for k in selected)

full = build(list(parts))
print("=== FULL PROMPT ===")
print(full)
print()
print("characters:", len(full))

# Now drop the two blocks beginners most often forget.
without = build([k for k in parts if k not in ("CONSTRAINTS", "FORMAT")])
print()
print("=== SAME PROMPT, NO FORMAT AND NO CONSTRAINTS ===")
print(without)
print()
print("characters:", len(without))
print()
print("Nothing here is magic: the second prompt still asks the question. What it")
print("stops doing is telling the model how long to be and what it may use, which")
print("is why replies to it drift in length and invent policy that does not exist.")

# 🎯 YOUR TURN — assemble a prompt from the six blocks.

parts = {
    "ROLE":        "You are a librarian at a city library.",
    "TASK":        "Answer the reader's question about borrowing.",
    "CONTEXT":     "Policy: adults may borrow 12 books for 3 weeks.",
    "FORMAT":      "Reply in one sentence.",
    "CONSTRAINTS": "Use only the policy text.",
}

# 1) Join each block as  "NAME:\n<text>"  with a blank line between blocks.
def build(selected):
    return "\n\n".join(f"{k}:\n{parts[___]}" for k in selected)
    # 👉 replace ___ with  k

# 2) Build the prompt from every block, in the order they appear above.
prompt = build(list(___))
# 👉 replace ___ with  parts

print(prompt)
print()
print("blocks:", len(prompt.split("\n\n")))

# ✅ Expected output:
# ROLE:
# You are a librarian at a city library.
#
# TASK:
# Answer the reader's question about borrowing.
#
# CONTEXT:
# Policy: adults may borrow 12 books for 3 weeks.
#
# FORMAT:
# Reply in one sentence.
#
# CONSTRAINTS:
# Use only the policy text.
#
# blocks: 5

Common Mistakes (and the fix)

• Stating the task but no format. — Ask for bullets, a table, or a set length.
• Forgetting constraints. — Name the limits (time, length, things to avoid).
• Skipping context. — Add the background only you know.
• Over-stuffing simple asks. — For easy tasks, a clear task line is enough.

⏱ Test Yourself — Timed Quiz

10 quick questions, 12 seconds each. Instant feedback — beat the clock!

🎉 Lesson Complete

• ✅ Six parts: role, task, context, format, constraints, examples
• ✅ Role and task set who acts and what they do
• ✅ Context informs; constraints set limits
• ✅ Format and examples shape the final answer
• ✅ Use only the parts your task actually needs

Practice quiz

Continue this course

• Previous: How AI Models Read Your Prompt
• Next: Context & Being Specific — Replace vague requests with the background and detail the model needs

Frequently asked questions

What are the parts of a great prompt?

A strong prompt usually combines six building blocks: a role (who the AI should act as), the task (what to do), context (background it needs), format (the shape of the answer), constraints (rules and limits), and examples (a sample of the style you want). You rarely need all six at once, but the more ambiguous your request, the more of them help.

Do I have to include every part every time?

No. Use the pieces that remove guesswork for your specific task. A quick question may only need a clear task. A trickier request, like rewriting copy for a specific audience in a specific format, benefits from adding role, context, format, and constraints. Think of them as ingredients you mix to taste.

What is the difference between context and constraints?

Context is the background the AI needs to do the job well, such as the audience, the situation, or your goal. Constraints are the rules and limits the answer must respect, such as a word count, a banned-jargon rule, or 'no new tools'. Context informs the task; constraints bound it.

Why do examples help so much?

An example shows the model the exact style, tone, or structure you want, which is far clearer than describing it in words. If you paste one good sample and say 'match this style', the AI can copy the pattern directly instead of guessing what 'good' looks like to you.