A prompt is a brief to a brilliant but contextless collaborator. Claude has read most of the public internet — but it doesn’t know:
| 🤔 Who you are | 🛠️ What you’ve tried | 🎯 What “good” means | 📦 What format you want |
|---|---|---|---|
| Your role | Past attempts | Your style | Length |
| Your audience | What failed | Your standards | Structure |
| Your industry | Constraints learned | Your taste | Tone |
Your job as the prompter is to close that gap fast. The single biggest improvement most people can make: stop writing one-line prompts and start writing 3–10 line briefs.
Almost every excellent prompt contains some mix of these six ingredients. Memorize them.
🎯 Great Prompt
│
┌───────────┬───────────┼───────────┬───────────┐
▼ ▼ ▼ ▼ ▼
🎭 ROLE 📋 CONTEXT 🎯 TASK 📐 CONSTRAINTS 📦 FORMAT 🌟 EXAMPLES
who Claude background the ask boundaries shape show, don't tell
is
| Ingredient | What it does | Mini-example |
|---|---|---|
| 🎭 Role | Sets perspective | “You are a senior product manager at a fintech startup.” |
| 📋 Context | Background info | “We’re launching a new card next month, target = young professionals.” |
| 🎯 Task | The actual ask | “Draft a 3-tier pricing page.” |
| 📐 Constraints | Boundaries | “Max 200 words per tier. No marketing fluff.” |
| 📦 Format | Output shape | “Return as a markdown table with: Tier | Price | Top 3 features.” |
| 🌟 Examples | Show, don’t tell | “Match how Stripe writes — clean, technical, confident.” |
[!TIP] Not every prompt needs all six. But if you find Claude’s output disappointing, look at this list and ask: which ingredient did I leave out?
┌─────────────────────────────────────────────────────────────────────┐
│ │
│ ❌ BEGINNER: "write me a cover letter for a PM job" │
│ │
│ ▼ produces ▼ │
│ │
│ 🤷 Generic Mad Libs template you can't actually send │
│ │
└─────────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────────┐
│ │
│ ✅ PRO: │
│ │
│ 🎭 You are an executive writing coach... │
│ │
│ 📋 <resume>...</resume> │
│ <job_description>...</job_description> │
│ <context>I met the hiring manager at a conference...</context>│
│ │
│ 🎯 Write a cover letter that: │
│ 📐 - Under 300 words │
│ 📐 - Opens with a specific hook (not "I'm writing to apply...") │
│ 📐 - Explicitly connects 2-3 resume bullets to the JD │
│ 📐 - References the conference meeting naturally │
│ 📐 - Sounds like a confident peer, not a desperate applicant │
│ │
│ 📦 Then in a section called "What I changed and why", │
│ explain your top 3 stylistic choices. │
│ │
│ ▼ produces ▼ │
│ │
│ ✨ Output you can actually send │
│ │
└─────────────────────────────────────────────────────────────────────┘
| # | Technique | One-liner |
|---|---|---|
| 1 | 🎯 Be specific about the audience | “Explain X to a smart 12-year-old who loves video games” |
| 2 | 📋 Show an example of “good” | Paste a sample. Claude matches tone & structure brilliantly |
| 3 | 👣 Break complex tasks into steps | Step 1: brainstorm. Step 2: outline. Step 3: draft |
| 4 | 🧠 Ask Claude to think step-by-step | Big quality boost on hard reasoning |
| 5 | 🏷️ Use XML tags for structure | Wrap inputs in <context>, <draft>, <task> |
| 6 | ✨ Few-shot examples | 2–3 input/output pairs, then ask for the next one |
| 7 | 🚫 Tell Claude what NOT to do | “no apologies, no disclaimers, no summary” |
| 8 | 🙋 Have Claude ask YOU first | “Ask me 3 clarifying questions before starting” |
| 9 | 🎯 Put role at the END | For long prompts, role last keeps it fresh |
| 10 | 🔧 Iterate, don’t restart | “Tighten paragraph 2” beats redoing the whole prompt |
❌ “Explain quantum computing.” ✅ “Explain quantum computing to a smart 12-year-old who loves video games. Use 1–2 analogies.”
When you have it, paste a sample. Claude is extraordinarily good at matching tone and structure when you show rather than describe.
Here's a tweet I wrote that I'm proud of:
"[paste tweet]"
Now write 5 more in the same voice on the topic of [X].
Instead of one giant prompt, walk Claude through a process:
We're going to write a blog post about [topic]. Let's do it in stages:
Step 1: Brainstorm 10 angles. I'll pick one.
Step 2: Build an outline with H2s and key points.
Step 3: Write the intro only. I'll review.
Step 4: Write the body section by section.
Start with Step 1.
This is called prompt chaining and it produces dramatically better long-form work.
For reasoning, math, or complex analysis, ask Claude to show its work before the answer:
Before giving your final answer, work through the problem step by step.
Then give the answer in a clearly labeled "Final answer:" line.
Claude was specifically trained to pay attention to XML-style tags. For prompts with multiple inputs, this is gold:
<context>
Our company sells handmade leather goods, founded in 2018, based in Portland.
</context>
<draft>
[paste your draft]
</draft>
<task>
Edit the draft above for clarity. Preserve the brand voice described in <context>.
Return only the edited version.
</task>
Convert customer messages into structured tickets.
Input: "My order hasn't arrived and it's been 2 weeks!"
Output: {"category": "shipping", "priority": "high", "sentiment": "negative"}
Input: "Loving the new feature, when's the next update?"
Output: {"category": "feedback", "priority": "low", "sentiment": "positive"}
Input: "Site is down, can't log in."
Output:
Before you start, ask me 3 clarifying questions that would
most change your approach.
This single trick will save you hours of bad output.
Counterintuitive but it works for long prompts. Putting the role/instruction at the end keeps it fresh in Claude’s “working memory”:
[long context...]
[example data...]
Given everything above, act as a forensic accountant and identify
the three most suspicious entries.
If output is 80% right, ask for surgical edits — don’t redo the whole prompt:
🚨 Bad output? Walk down this list — first match wins.
| If… | Then… |
|---|---|
| ❓ Task is ambiguous | Add: “Ask me 3 clarifying questions first” |
| 🎨 Wrong tone or format | Add: an example or a format spec |
| 🧩 Complex multi-step task | Break into a chain of smaller prompts |
| 🧠 Reasoning is shallow | Add: “Think step-by-step before answering” |
| 📋 Details are missing | Add: role + context + constraints |
| 🤷 Still bad? | Try a different model (Sonnet → Opus) |
Explain [concept] using:
- A 1-sentence ELI5
- A concrete real-world analogy
- The technical definition
- One common misconception people have
Keep the whole thing under 250 words.
You're an experienced [role]. Below is my [draft/code/plan].
<work>
[paste]
</work>
Give me:
1. Three things working well (be specific, not generic)
2. Three things to improve, ranked by impact
3. One bold suggestion I probably haven't considered
Be direct. I'd rather be helped than flattered.
I'm deciding between [Option A] and [Option B]. Context:
- [your situation]
- [constraints]
- [what success looks like]
Don't just list pros/cons. Tell me:
1. Which option you'd pick and why
2. The one scenario where you'd flip your answer
3. What I should learn or test before deciding
Push back on my framing if you think I'm asking the wrong question.
┌──────────────────────────────────┬──────────────────────────────────┐
│ ❌ MISTAKE │ ✅ FIX │
├──────────────────────────────────┼──────────────────────────────────┤
│ One-line prompts │ Add role + context + format │
│ Yes/no on complex topics │ Ask for tradeoffs + conditions │
│ Stuffing 10 tasks in one prompt │ Split into a chain │
│ Vague feedback ("make it nice") │ Be specific: "shorter,formal" │
│ Re-explaining context every turn│ Use a Project (Module 04) │
│ Accepting first answer │ Ask "what would 50% better be?" │
└──────────────────────────────────┴──────────────────────────────────┘
The best prompters share three habits — they all lead to the same place:
| Habit | Why it works |
|---|---|
| 🧪 Run experiments, not requests | Every prompt is a hypothesis to test |
| 📚 Keep a personal prompt library | Your best prompts are an asset; save them |
| 🔍 Study the model’s failures | Bad outputs teach you what’s missing |
→ All three roads lead to 🎯 consistently great output.
When Claude misfires, ask “what was missing in my prompt?” — not “why is the AI dumb?”
You should now be able to:
[!IMPORTANT] 🏋️ Exercise: Take a prompt you used last week and rewrite it using everything from this module. Compare the outputs side-by-side.
👉 Next up: Module 04 — Claude.ai Features — Projects, Artifacts, Files, Memory, and Connectors.
| ← Previous | 🏠 Home | Next → |
|---|---|---|
| Module 02 — Getting Started | Course README | Module 04 — Claude.ai Features |