AI Prompt Generator — Transform Raw Text Into Structured Prompts

How the AI Prompt Generator Works

The prompt generator uses a fine-tuned 7B model for organizing raw text into the six-field sinc format. Related: sinc-LLM AI prompt generator.

The author uses a signal-processing analogy to discuss natural-language specification:

The format separates PERSONA, CONTEXT, DATA, CONSTRAINTS, FORMAT, and TASK. Review the proposed fields and verify the response against the task and supplied sources.

What Makes This Different From Other Prompt Generators

Most AI prompt generators use static templates: "Act as a [role], I want you to [task], output as [format]." These templates cover 2-3 specification dimensions at most, leaving the LLM to guess the rest.

The sinc-LLM approach is fundamentally different. Instead of filling templates, the AI analyzes what specification dimensions are present in your raw text, what dimensions are missing, and what content should fill those gaps based on the context of your request.

For example, if you write "help me write a cover letter for a data science job," the AI prompt generator recognizes:

• Implicit PERSONA: Professional career coach / HR expert
• Implicit CONTEXT: Job application, data science field, competitive market
• Missing DATA: The user's background, target company, specific role requirements
• Missing CONSTRAINTS: Length, tone, which achievements to highlight
• Missing FORMAT: Business letter format, sections, paragraph structure
• Explicit TASK: Write a cover letter

The generator fills in reasonable defaults for missing bands while flagging them so you can customize. This means every generated prompt is complete — no specification gaps for the LLM to fill with guesses.

Example: AI-Generated Structured Prompt

Input: "Write a Python script that scrapes product prices from Amazon"

{
  "formula": "x(t) = Σ x(nT) · sinc((t - nT) / T)",
  "T": "specification-axis",
  "fragments": [
    {
      "n": 0,
      "t": "PERSONA",
      "x": "Senior Python developer specializing in web scraping and data extraction. Pragmatic, security-conscious, aware of legal and ethical considerations."
    },
    {
      "n": 1,
      "t": "CONTEXT",
      "x": "Building a price monitoring tool for personal use. Amazon uses dynamic rendering and anti-bot measures. Need a solution that respects robots.txt and rate limits."
    },
    {
      "n": 2,
      "t": "DATA",
      "x": "Target: Amazon product pages. Extract: product title, current price, original price, discount percentage, availability status. Input: list of ASINs or product URLs."
    },
    {
      "n": 3,
      "t": "CONSTRAINTS",
      "x": "Use requests + BeautifulSoup (no Selenium unless necessary for JS rendering). Include User-Agent rotation. Add 2-5 second random delays between requests. Handle CAPTCHAs gracefully (skip and log, do not solve). Output to CSV. Include error handling for 503 responses. Do not use proxies. Include robots.txt check. Python 3.10+. No async."
    },
    {
      "n": 4,
      "t": "FORMAT",
      "x": "Single Python file with clear function separation. Docstrings on all functions. CSV output with headers. Print progress to stdout."
    },
    {
      "n": 5,
      "t": "TASK",
      "x": "Write the complete Python scraping script following all specifications above."
    }
  ]
}

Notice how the AI generator added critical specification that was completely absent from the raw input: rate limiting, error handling, legal considerations, output format, and library choices. Without these bands, the LLM would make arbitrary decisions about each dimension.

Use Cases for the AI Prompt Generator

Rapid prototyping: When you know what you want but do not have time to write a detailed prompt, paste your rough idea and let the sinc-LLM AI prompt generator structure it for you.

Prompt library building: Generate structured prompts for common tasks, save them as .sinc.json files, and reuse them across projects. The JSON format is version-controllable and diffable.

Teaching prompt engineering: Show students how a raw prompt maps to the 6-band structure. The generated output makes the implicit explicit, which is the core skill of prompt engineering.

Multi-model testing: Generate one structured prompt, then test it across ChatGPT, Claude, Gemini, and Grok. The 6-band format is model-agnostic, so you get a fair comparison of model capabilities rather than model sensitivity to prompt phrasing.

Agent system development: The sinc JSON format serves as the inter-agent communication contract in multi-agent architectures. Generate the initial prompt structure, then let agents pass sinc JSON to each other with full specification fidelity.

Why 6 Bands? The Signal Theory Behind Prompt Generation

The number 6 is not arbitrary. Research into LLM specification failures revealed that prompts fail along exactly 6 orthogonal dimensions: identity (who), background (what surrounds the task), input (what to work with), rules (what to avoid), structure (how to output), and action (what to do). These map to PERSONA, CONTEXT, DATA, CONSTRAINTS, FORMAT, and TASK respectively.

The sinc-LLM prompt generator treats these 6 dimensions as the Nyquist frequency of prompt specification. Sample at this rate and you capture the full bandwidth of human intent. Sample below it (fewer than 6 bands) and you get aliasing — the LLM fills gaps with assumptions that may not match your intent.

This is why the CONSTRAINTS band (n=3) is always the longest. Constraints carry the most specification density — they are the high-frequency components of your intent. A prompt without explicit constraints is like an audio signal sampled at too low a rate: it sounds wrong, and you cannot fix it by amplifying the signal.