# Prompt engineering basics

## 🎯 Goal

Learn how to write better prompts to get the most out of ChatGPT and similar tools — for *all* kinds of tasks, not just coding.

***

## 🚀 Why This Matters

Most people write prompts like this:

> “Explain X”\
> “Write code for Y”\
> “What’s a good name for a feature?”

That’s the equivalent of Googling "stuff".

Let’s go pro.

***

## 📦 Core Concepts of Prompting

### 1. **Context is King**

LLMs love context. The more you give, the better the output.

> ❌ “Write a blog post about AI.”\
> ✅ “Write a 500-word blog post about how small dev agencies like MarsBased can use AI to automate documentation. Audience: CTOs. Tone: witty but professional.”

> ✅ “You're a senior backend dev at MarsBased. Given this GitHub issue, propose 3 implementation options with tradeoffs.”

***

### 2. **Role Assignment**

> “You're a senior full-stack dev.”\
> “Act like a sassy but accurate technical recruiter.”\
> “You're a product manager writing internal release notes.”

Roles change everything. Try it.

***

### 3. **Examples > Explanations**

Few-shot prompting is powerful.

> ❌ “Write me an error message.”\
> ✅ “Here are 2 examples of error messages from our system:\n\n1. 'We couldn’t connect to GitHub. Check your token.'\n2. 'Your session expired. Please log in again.'\n\nNow write a message for a failed API sync with Linear.”

***

### 4. **Constraints Help**

> “Explain this like I’m five.”\
> “Summarize this in 3 bullet points.”\
> “Reply in markdown, use emojis.”\
> “Write code with comments in Spanish.”

Constraints = better outputs.

***

### 5. **Iterate, Don’t Expect Magic**

You’re allowed to keep going:

> “That’s too long. Can you shorten it?”\
> “Add more sarcasm.”\
> “Use TypeScript instead.”\
> “Rewrite using a more optimistic tone.”

LLMs love feedback.

***

## 🔧 Dev Use Cases

* 🔍 Debugging:\
  Paste an error and ask for causes + fixes.
* 📄 PR Summaries:\
  “Summarize this PR in 3 sentences for a non-dev.”
* 🧠 Naming things:\
  “Suggest 10 function names. Context:…”
* 📜 Commit messages:\
  “Write a conventional commit for this diff.”
* 🤖 GitHub Copilot prompts:\
  Use natural language to write comments that guide code generation.

***

## 💼 Non-Dev Use Cases

* ✍️ Docs:\
  “Turn this bullet list into internal documentation.”\
  “Write a changelog based on these commits.”
* 📣 Marketing:\
  “Write a tweet thread from this blog post.”\
  “Summarize this podcast transcript for LinkedIn.”
* 📅 Planning:\
  “Create a rough roadmap based on these 4 Linear issues.”
* 🧑‍🏫 Hiring:\
  “Generate 5 technical interview questions for a senior Vue dev.”

***

## ⚙️ Bonus: API-Level Prompting (1 Minute)

If using OpenAI's API:

```json
{
  "model": "gpt-4",
  "temperature": 0.7,
  "messages": [
    { "role": "system", "content": "You are a senior Rails engineer." },
    { "role": "user", "content": "Explain the pros and cons of Hotwire." }
  ]
}
```

* **system**: Sets the role/persona
* **user**: The actual prompt
* **temperature**: 0 = deterministic, 1 = creative/random

## 🎁 Cheat Code: Prompt Structure

* \[Persona or role]
* \[Clear task]
* \[Context]
* \[Constraints]
* \[Optional: examples]
* \[Optional: output formatting]

🔥


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://handbook.marsbased.com/our-guides/prompts.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.