▶ Click to watch the full video

Prompt Improver

An interactive CLI tool that transforms simple prompts into comprehensive, professional-quality prompts using local LLMs. Built on the 8 Pillars of Effective Prompts framework.

Simple prompts produce simple outputs. Structured prompts produce professional intelligence.

What It Does

Paste in a basic prompt like:

Show assets missing SentinelOne

Get back a comprehensive prompt that produces 19-page professional reports instead of 8-page data dumps -- same AI, same data, only the prompt changed.

The 8 Pillars Framework

#	Pillar	What It Adds
1	Specificity & Context	Scope, audience, domain boundaries
2	Structure & Organization	Executive summary, methodology, findings, recommendations
3	Data Requirements	Tables, metrics, percentages, breakdowns
4	Risk & Impact	Risk matrices, business impact, compliance
5	Timeline Actions	Immediate / short-term / long-term recommendations
6	Quality Elements	Data sources, validation, limitations
7	Accountability	Ownership, KPIs, success criteria
8	Professional Formatting	Headers, tables, confidentiality labels

Prerequisites

• Bash 4.0+ (macOS/Linux)
• Ollama -- local LLM inference
• curl and jq -- usually pre-installed

Quick Start

# 1. Install Ollama (if not already installed)
curl -fsSL https://ollama.com/install.sh | sh

# 2. Pull the default model
ollama pull gemma3:12b

# 3. Start Ollama (if not already running)
ollama serve &

# 4. Clone this repo
git clone https://github.com/YOUR_USERNAME/PromptImprover.git
cd PromptImprover

# 5. Make executable and run
chmod +x improve-prompt.sh
./improve-prompt.sh

Usage

Interactive Mode (default)

./improve-prompt.sh

Type prompts at the > prompt. The tool will:

1. Analyze intent -- detect if your prompt is vague and ask clarifying questions
2. Enhance -- apply the 8 Pillars framework
3. Display -- show the improved prompt with word count stats

Single Prompt

./improve-prompt.sh "Generate a security report"

From File

./improve-prompt.sh -f my-prompt.txt
./improve-prompt.sh -f my-prompt.txt -o enhanced-prompt.txt

Options

Flag	Description
-h, --help	Show help
-f, --file FILE	Read prompt from file
-o, --output FILE	Save enhanced prompt to file
-m, --model MODEL	Use a different Ollama model
--no-intent	Skip intent analysis (clarifying questions)
--no-logo	Skip the startup logo animation

Interactive Commands

Command	Description
/help	Show available commands
/intent	Toggle intent analysis on/off
/model	Switch to a different model
/history	Show recent prompt history
/save	Save last enhanced prompt to file
/clear	Clear the screen
/quit	Exit

Using a Different Model

Any model available through Ollama works:

# Use a smaller model for faster results
./improve-prompt.sh -m gemma3:4b "List vulnerabilities"

# Use a larger model for better quality
./improve-prompt.sh -m llama3:70b "Generate a compliance report"

# Switch models in interactive mode
/model

Example: Before & After

Before (simple prompt):

Generate a report showing assets missing SentinelOne coverage

After (enhanced prompt):

Generate a comprehensive Asset Inventory Gap Analysis for endpoints missing SentinelOne agent coverage, intended for IT Security Management and senior leadership.

Include: Executive Summary (under 200 words), Methodology (data sources, validation, exclusions), Findings (breakdown by OS, criticality, business unit with percentages), Risk Assessment (impact matrix for each category), and Recommendations (immediate 0-30 days, short-term 30-90 days, long-term 90-180 days with assigned ownership).

Format as professional report with tables, clear headers, and confidentiality label. Target length: 15-20 pages.

Result: 19-page professional analysis vs. 8-page basic list.

Project Structure

PromptImprover/
├── improve-prompt.sh        # Main CLI tool
├── zafran-logo.sh           # Animated ASCII logo
├── zafran-logo-static.sh    # Static ASCII logo
├── training/                # Workshop materials
│   ├── TRAINING_OUTLINE.md
│   ├── QUICK_REFERENCE_CARD.md
│   ├── SPEAKER_NOTES.md
│   └── DESIGN_GUIDE.md
├── Short Prompt.pdf         # Example: basic prompt output
├── Long Prompt.pdf          # Example: enhanced prompt output
├── LICENSE
└── README.md

Configuration

The tool uses sensible defaults but can be customized via environment variables:

# Use a different Ollama endpoint
export OLLAMA_URL="http://localhost:11434/api/generate"

# Skip the logo on startup
export SHOW_LOGO=false

Contributing

Contributions are welcome! Here are some ways to help:

• Report bugs -- open an issue describing the problem
• Suggest enhancements -- ideas for new pillars, better system prompts, or UX improvements
• Add model support -- test and document results with different Ollama models
• Improve training materials -- expand examples, add new use cases, translate to other languages
• Share your results -- before/after comparisons from your own use cases

How to Contribute

1. Fork the repository
2. Create a feature branch (git checkout -b feature/my-improvement)
3. Make your changes
4. Test with ./improve-prompt.sh to verify everything works
5. Submit a pull request

License

This project is licensed under the MIT License -- see LICENSE for details.

The MIT License covers the code, prompt templates, and training materials. You are free to use, modify, and distribute everything in this repository, including the prompt engineering framework and system prompts, for any purpose.