Capstone: Build an AI Writing Assistant with Prompt Engineering

Project Overview

In this capstone, you’ll build a complete AI writing assistant using only prompt engineering techniques — no fine-tuning, no RAG.

The assistant will:

• Accept a topic and target audience
• Generate structured content (blog post, email, report)
• Enforce a consistent brand voice
• Return structured JSON with the content and metadata

System Prompt Design

You are a professional content strategist and copywriter.

You write clear, compelling content tailored precisely to the target audience.
You never use filler phrases, passive voice, or unnecessary adjectives.
You always follow the requested format exactly.

Output format: Always return valid JSON matching the schema provided.

Step 1: Content Brief Extractor

Build a prompt that extracts a structured brief from a user’s rough description:

brief_prompt = """
Extract a content brief from the user's description. Return JSON:
{
  "topic": "string",
  "audience": "string",  
  "tone": "professional|casual|technical|inspirational",
  "format": "blog|email|report|social",
  "key_points": ["string"]
}

User description: {user_input}
"""

Step 2: Content Generator

Use the brief to generate the content:

content_prompt = """
Write a {format} about {topic} for {audience}.

Key points to cover:
{key_points}

Tone: {tone}
Length: {length_guidance}

Return JSON: { "title": string, "content": string, "word_count": number }
"""

Step 3: Quality Checker

Add a final evaluation step that scores the output before returning it to the user.

Extension Challenge

Add a few-shot example bank — 3 examples of high-quality outputs in your target style — to the system prompt to dramatically improve consistency.