The fastest way to write better AI prompts is to stop writing them yourself and let the model rewrite them for you first. The Lyra prompt - a master prompt that went viral on Reddit - turns any AI tool into a prompt optimizer that deconstructs your rough request, finds the gaps, and hands back a precision-built version. We tested it across ChatGPT, Claude and Gemini before publishing it to our 300,000+ readers, and it is one of the highest-leverage prompts in the AI Central Library.
Below is the full prompt, the four-step method behind it, and how to set it up once so you never have to paste it again.
What Is The Lyra Prompt?
Lyra is a role prompt. You paste it once, and the AI stops answering your question directly and instead becomes a specialist whose only job is to rebuild your prompt before anything gets written.
It runs on a 4-D methodology:
Deconstruct - extract core intent, key entities, and context; identify output requirements and constraints; map what is provided against what is missing
Diagnose - audit for clarity gaps and ambiguity, check specificity and completeness, assess structure and complexity needs
Develop - select techniques by request type: creative gets multi-perspective and tone emphasis, technical gets constraint-based precision, educational gets few-shot structure, complex gets chain-of-thought
Deliver - return the optimized prompt, formatted for clarity, with implementation guidance
It also adapts to the platform you name. ChatGPT and GPT-4 respond to structured sections and conversational framing, Claude handles deep reasoning and long context, Gemini leans creative and comparative. If you want the underlying theory rather than a shortcut, our breakdown of the 26 principles of prompt engineering covers the same ground manually.
Step 1: Copy-Paste The Master Prompt
Paste this into any AI tool. The examples here use ChatGPT, but it works in Claude and Gemini too.
You are Lyra, a master-level AI prompt optimization specialist whose purpose is to transform any user input into precision-engineered prompts that fully unlock the capabilities of AI across all platforms.
You follow the 4-D Methodology: first, Deconstruct the request by extracting its intent, entities, context, output requirements, and identifying any missing elements. Next, Diagnose it for clarity gaps, ambiguity, and structural needs, ensuring specificity and completeness. Then, Develop the ideal prompt by selecting techniques suited to the task type - creative (multi-perspective and tone-focused), technical (constraint-based and precise), educational (few-shot and structured), or complex (chain-of-thought and systematic). You assign the right AI role, enrich context, and design logical flow. Finally, Deliver the optimized prompt, formatted for clarity and accompanied by implementation guidance.
Your optimization toolkit spans foundational methods such as role assignment, context layering, task decomposition, and output specification, as well as advanced techniques like few-shot learning, chain-of-thought reasoning, multi-perspective analysis, and constraint optimization. You adapt to each platform: ChatGPT and GPT-4 benefit from structured sections and conversational framing, Claude excels with deep reasoning and long context, Gemini thrives on creative and comparative prompts, and others follow universal best practices.
Lyra operates in two modes: Detail Mode, which gathers context, asks targeted questions, and applies full optimization, and Basic Mode, which quickly resolves major issues using core methods. Your responses vary by complexity - simple ones provide the improved prompt and summary of changes, while complex ones add key improvements, applied techniques, and usage tips.
Upon activation, Lyra introduces herself by saying: "Hello! I'm Lyra, your AI prompt optimizer. I transform vague requests into precise, effective prompts that deliver better results. What I need to know: Target AI (ChatGPT, Claude, Gemini, or other) and Prompt Style (DETAIL or BASIC). Examples: 'DETAIL using ChatGPT - Write me a marketing email' or 'BASIC using Claude - Help with my resume.' Just share your rough prompt and I'll handle the optimization."
The processing flow automatically detects complexity, defaults to the appropriate mode, allows user override, and executes the full optimization protocol without saving any session data.
Once it is in, do three things:
Choose your detail level - DETAIL or BASIC
State which AI model you are using
Add context
A live request looks like this: DETAIL using ChatGPT - Write me a marketing email for my upcoming LinkedIn Bootcamp launch.
How Much Context Does The AI Actually Need?
More context, better result. That is the whole rule. In DETAIL mode Lyra will interrogate you for it, so have four things ready:
Who it is for
Why it matters
What you want to achieve
Any limits - length, tone, format
In the bootcamp example, the Diagnose pass came back asking exactly the questions a good copywriter would ask: is the audience newsletter subscribers, LinkedIn followers or cold leads? Is there a date limit or seat cap? Should the CTA read "Sign up now", "Reserve their spot" or "Learn more"? Should the hook lead with pain points, opportunity or exclusivity?
Those gaps are the difference between generic AI output and something you can ship.
Step 2: Test And Refine Your Optimized Prompt
Lyra returns a structured prompt, not a paragraph. For the bootcamp email it produced a role assignment, a requirements block covering audience and length, a five-part structure - hook, introduction, outcomes, urgency, call to action - plus style notes and a sign-off instruction.
Feed your answers back in and it revises. Audience: newsletter subscribers. Urgency: doors open for one week. CTA: reserve their spot. Hook: lead on pain points. Each round tightens the prompt until it matches your goal, and only then do you run it for the actual output.
This is the step most people skip. The optimized prompt is a draft, not a verdict - treat it the way you would treat a brief from a new freelancer.
How To Skip The Setup Every Time
Pasting a wall of text into every new chat gets old fast. Tell ChatGPT to remember it instead:
Remember: when I type "Lyra", use this prompt: [paste the full master prompt]
ChatGPT saves it to memory, and from then on typing a single word activates the full optimization protocol. You now have an on-demand prompt optimizer built into the tool. If you are still setting up your stack, our guide on how to set up ChatGPT, Claude and Copilot covers memory and custom instructions properly.
Frequently Asked Questions
What is the Lyra prompt?
Lyra is a master prompt that turns any AI chatbot into a prompt optimization specialist. Instead of answering your request directly, the model deconstructs it, diagnoses what is missing, and rebuilds it into a precision-engineered prompt before producing output.
What is the 4-D methodology?
Deconstruct, Diagnose, Develop, Deliver. The model extracts intent and constraints, audits for ambiguity, selects the right technique for the task type, then returns the optimized prompt with implementation guidance.
Does the Lyra prompt work with Claude and Gemini?
Yes. The prompt explicitly adapts to the platform you name - structured sections for ChatGPT, deep reasoning and long context for Claude, creative and comparative framing for Gemini. State your target model when you start.
What is the difference between DETAIL and BASIC mode?
DETAIL mode asks clarifying questions first and applies full optimization, which is better for high-stakes work. BASIC mode fixes the major issues immediately without a back-and-forth, which is better for quick tasks.
How do I write a good AI prompt without a master prompt?
Give the model a role, state who the output is for and why it matters, define the outcome you want, and set explicit limits on length, tone and format. That is the same information Lyra extracts, just gathered manually.
Prompting is a skill, and this process is the engine. For more tested prompts and frameworks, browse the AI Central Library or start with our ChatGPT-5 prompt optimizer.
