The Constraint Specification: A Technical Standard for Prompt Quality

The Need for Standards

In the 1960s, software engineers had no coding standards. Each programmer wrote code in their own way. The code was messy and hard to fix. Then the industry built standards: structured programming, design patterns, SOLID principles, code reviews, and testing. Quality rose. Costs fell. The industry grew up.

AI prompting in 2026 is where software was in the 1960s. Each user writes prompts in their own way. There is no standard for what makes a prompt complete. There is no check for prompt quality. Nothing stands between a bad prompt and the model. The result: outputs that vary, high hallucination rates, wasted tokens, and billions in failed projects.

The AI industry needs what software engineering got: a formal standard for inputs. That is why I built one.

The sinc-prompt Specification

My sinc-prompt specification is a formal standard for writing prompts. It defines:

• 6 mandatory specification bands (PERSONA, CONTEXT, DATA, CONSTRAINTS, FORMAT, TASK) with set quality weights.
• A JSON schema for encoding prompts in a structured way (JSON Schema draft-07).
• Validation rules to check that all 6 bands are present with enough tokens in each.
• Quality metrics (SNR score, zone function check, band coverage score).
• An interactive validator at sincllm.com/validate to check your prompt quality in real time.

Mandatory Fields and Validation

The specification needs all of the following:

{
  "formula": "x(t) = Sigma x(nT) * sinc((t - nT) / T)",
  "T": "[domain]-axis",
  "fragments": [
    {
      "n": 0,
      "t": "PERSONA",
      "x": "[WHO answers \u2014 role, expertise, voice]"
    },
    {
      "n": 1,
      "t": "CONTEXT",
      "x": "[SITUATION \u2014 background, environment, prior state]"
    },
    {
      "n": 2,
      "t": "DATA",
      "x": "[INPUTS \u2014 specific facts, numbers, references]"
    },
    {
      "n": 3,
      "t": "CONSTRAINTS",
      "x": "[RULES \u2014 boundaries, prohibitions, requirements \u2014 minimum 40% of total tokens]"
    },
    {
      "n": 4,
      "t": "FORMAT",
      "x": "[STRUCTURE \u2014 output shape, sections, types]"
    },
    {
      "n": 5,
      "t": "TASK",
      "x": "[OBJECTIVE \u2014 what to produce]"
    }
  ]
}

Validation rules:

1. All 6 fragments must be present (n=0 through n=5).
2. Each fragment must have a non-empty x field.
3. The CONSTRAINTS fragment (n=3) must be at least 40% of the total token count.
4. Total prompt tokens should be in the 209-233 range (you get a warning if outside this range).
5. The formula and T fields must be present.

Quality Metrics

The specification defines 3 quality metrics:

1. SNR (Signal-to-Noise Ratio): The share of tokens that do real work, out of all tokens. Target: ≥ 0.70. To compute it, count the tokens that cut model uncertainty across any band, then divide by the total token count.

2. Band Coverage: The share of the 6 bands that hit their minimum token count. Target: 100% (6/6). Each band has a minimum: PERSONA (10 tokens), CONTEXT (15 tokens), DATA (10 tokens), CONSTRAINTS (40 tokens), FORMAT (15 tokens), TASK (5 tokens).

3. Zone Function Score: A weighted quality score across 4 zones (Z1: identity, Z2: information, Z3: constraints, Z4: format). It comes from the 1M-simulation research. Target: ≥ 0.70 composite.

Adoption Path

The specification is built for step-by-step adoption:

1. Level 1: Awareness. Use the validator to check your current prompts. Find where the gaps are.
2. Level 2: Manual structuring. Add the missing bands by hand to your most important prompts.
3. Level 3: Tool integration. Use sinc-LLM to split raw prompts into 6-band signals automatically.
4. Level 4: Pipeline integration. Add sinc validation to your AI pipeline as a quality gate. Block prompts with an SNR below 0.70.
5. Level 5: Organizational standard. Make sinc-prompt the standard prompt format for your whole team or company.

The specification is open (MIT license), easy to extend, and works with any model. It works with GPT-4, Claude, Gemini, Llama, and any future LLM because it targets the input layer, not the model layer.

The AI industry has spent 4 years making models better. Now it is time to standardize the inputs. The era of signal design starts with a formal specification.