Skip to main content
Menu
Context is King

P.R.O.M.P.T.

A Principle-Based Approach for LLM-Human Communication

Transform prompt engineering from ad-hoc guessing to systematic methodology. Any pattern can (and should) be broken down to principles.

6
Components
1
Core Principle
5+
Domains
Applications

The Six Principles

Click on any component to explore its purpose, examples, and best practices.

Click on any letter above to explore that component in detail

Interactive Prompt Builder

Build your prompt step-by-step using all six components. Watch it come together in real-time.

Prompt Completeness0/6 components

Assembled Prompt

Start filling in the components on the left to build your prompt...

PROMPT

Try It Live

Transform any prompt using the P.R.O.M.P.T framework — paste your prompt, get back a structured version.

Launch P.R.O.M.P.T Bot

Powered by GPT-4 • Free to use

Real-World Examples

See P.R.O.M.P.T. in action across different domains.

🔐

Security Assessment

Cybersecurity
P: Create a vulnerability assessment report for our Q1 security audit
R: Markdown table with CVSS scores, followed by prioritized remediation steps
O: Do not include theoretical vulnerabilities; only confirmed findings from our scan results
M: Audience: CISO and security team familiar with NIST CSF
P: Professional tone, executive summary first, technical details in appendix
T: Use CVSS 3.1, reference CVE IDs, map to MITRE ATT&CK where applicable
View assembled prompt
You are a senior security analyst preparing a Q1 vulnerability assessment for the CISO and security team (familiar with NIST CSF).

Create a vulnerability assessment report with:
- Executive summary (2-3 paragraphs)
- Markdown table: Vulnerability | CVSS 3.1 Score | CVE ID | MITRE ATT&CK Mapping | Status
- Prioritized remediation steps

Guidelines:
- Only include confirmed findings from scan results (no theoretical vulnerabilities)
- Professional tone throughout
- Technical details in appendix section
- Use CVSS 3.1 scoring methodology
              
📝

Technical Blog Post

Content Creation
P: Write a blog post explaining prompt injection attacks for developers
R: Long-form article with code examples, diagrams descriptions, and actionable takeaways
O: Avoid jargon without explanation; do not provide actual exploit code
M: Readers are web developers new to AI security, learning about LLM integration risks
P: Conversational but authoritative tone, 1500-2000 words, use analogies
T: Reference OWASP LLM Top 10, include Python/JavaScript pseudocode for defenses
View assembled prompt
Write a technical blog post about prompt injection attacks for web developers who are new to AI security but experienced in web development.

Structure:
- Engaging introduction with real-world analogy
- What is prompt injection (with simple examples)
- Types of attacks (direct vs indirect)
- Code examples showing vulnerable vs secure patterns (Python/JavaScript pseudocode)
- Defense strategies with implementation guidance
- Key takeaways section

Guidelines:
- 1500-2000 words
- Conversational but authoritative tone
- Explain jargon when first introduced
- Reference OWASP LLM Top 10
- No actual exploit code; focus on defense patterns
- Include spots for diagrams with descriptions
              
👨‍💻

Code Review

Development
P: Review this authentication module for security vulnerabilities and code quality
R: Categorized findings: Critical/High/Medium/Low with line references and fix suggestions
O: Focus only on security and maintainability; ignore styling/formatting issues
M: This is a production Node.js API handling financial transactions
P: Direct and specific feedback, code snippets for suggested fixes
T: Check against OWASP ASVS, consider GDPR data handling requirements
View assembled prompt
Review the following Node.js authentication module for a production financial transactions API.

Focus areas:
- Security vulnerabilities (OWASP ASVS compliance)
- GDPR data handling requirements
- Code maintainability

Output format:
| Severity | Line | Issue | Recommendation |
|----------|------|-------|----------------|
| Critical | ... | ... | ... |

For each finding:
- Specific line reference
- Clear explanation of the risk
- Code snippet showing the fix

Ignore: Code styling, formatting, naming conventions (handled by linter)
              
🔬

Research Summary

Academic
P: Summarize recent advances in adversarial machine learning for my literature review
R: Structured summary with themes, key papers, and identified research gaps
O: Only include peer-reviewed sources from 2022-2025; clearly mark any preprints
M: I am a PhD candidate writing a dissertation chapter on AI robustness
P: Academic writing style, proper citation format (APA 7), balanced perspective
T: Cover evasion attacks, poisoning attacks, and certified defenses; include mathematical notation where relevant
View assembled prompt
Summarize recent advances (2022-2025) in adversarial machine learning for a PhD dissertation literature review chapter on AI robustness.

Structure:
1. Overview of the field evolution (2022-2025)
2. Key themes:
   - Evasion attacks and defenses
   - Data poisoning attacks
   - Certified robustness methods
3. Seminal papers table: Paper | Authors | Year | Key Contribution | Citation count
4. Identified research gaps
5. Future directions

Guidelines:
- Academic writing style
- APA 7 citation format
- Only peer-reviewed sources; clearly mark preprints with [preprint]
- Include mathematical notation for key concepts
- Balanced perspective acknowledging limitations
              

The Complete Picture

Two frameworks, one unified AI security methodology.

⚔️

AATMF

Offensive Framework

How attackers exploit LLM systems. 20 tactics, 240+ techniques for understanding adversarial AI threats.

Explore AATMF →
🛡️

P.R.O.M.P.T.

Defensive Framework

How to communicate effectively with LLMs. 6 principles for precise, reliable, and safe AI interactions.

You are here

Understanding both frameworks creates a complete picture: knowing how attacks work (AATMF) helps you design prompts that are both effective AND resistant to manipulation (P.R.O.M.P.T.).