A meta-prompt is a prompt that instructs an LLM to generate other prompts. sinc-LLM's meta prompt generator creates 6-band meta-prompts that produce structured, specification-complete prompts for any task — recursively improving output quality at every level.
A meta-prompt operates one level of abstraction above a regular prompt. Instead of asking the LLM to "write a blog post," a meta-prompt asks the LLM to "generate a complete prompt specification for writing a blog post." The output is not the blog post — it is a better prompt that will produce a better blog post.
The sinc-LLM meta prompt generator takes this concept further by ensuring every meta-prompt produces output in the 6-band sinc JSON format. This means the generated prompts are not just "better" in a vague sense — they are structurally complete, with all 6 specification dimensions specified at the Nyquist rate.
The gap between what you know you want and what you can specify is the prompt engineering gap. You know you want "a good marketing email," but translating that into a complete specification — persona, context, data, constraints, format, task — requires expertise that most users do not have.
Meta-prompts bridge this gap. You describe your goal at a high level, and the meta-prompt generates the detailed specification. It is like having a prompt engineer on staff who translates your vague requirements into precise specifications.
In the sinc-LLM framework, a meta-prompt ensures that the generated specification covers all 6 frequency bands. Without a meta-prompt, the LLM generates prompts with the same gaps and biases that plague raw human prompts. With a sinc-structured meta-prompt, every generated prompt is specification-complete.
Here is what a sinc-LLM meta-prompt looks like in sinc JSON format:
{
"formula": "x(t) = \u03a3 x(nT) \u00b7 sinc((t - nT) / T)",
"T": "specification-axis",
"fragments": [
{"n": 0, "t": "PERSONA", "x": "Expert prompt engineer specializing in sinc-LLM 6-band decomposition"},
{"n": 1, "t": "CONTEXT", "x": "User needs prompts for generating product descriptions for a Shopify e-commerce store selling handmade ceramics"},
{"n": 2, "t": "DATA", "x": "Product categories: mugs, bowls, plates, vases. Price range: $25-$150. Target: home decor enthusiasts aged 28-45"},
{"n": 3, "t": "CONSTRAINTS", "x": "Each generated prompt MUST include all 6 sinc bands. PERSONA must be a copywriter with e-commerce experience. CONSTRAINTS must specify SEO keywords, character limits (150-300 chars for meta, 500-800 words for description), brand voice (warm, artisanal, authentic), and must mention materials and dimensions. FORMAT must request HTML with schema.org Product markup. Generated prompts must be reusable across product categories with minimal modification."},
{"n": 4, "t": "FORMAT", "x": "Output as valid sinc JSON with all 6 bands. One prompt per product category."},
{"n": 5, "t": "TASK", "x": "Generate 4 complete sinc-LLM prompts, one for each product category, that will produce SEO-optimized product descriptions"}
]
}
sinc-LLM makes meta-prompting accessible. You do not need to understand the theory — the tool generates the meta-prompt structure automatically. But understanding why it works gives you an edge: the 6-band decomposition is not arbitrary. It is the mathematically optimal sampling rate for LLM task specification.