Engineering reference
Practices for Structured Prompt Preparation
The author's structured-prompt practices, including explicit fields, constraints, supplied data, personas, and output formats.
The services, methods, and examples described here do not establish an expected performance result for your workload.
By Mario Alexandre · March 27, 2026 · 10 min read
These practices describe the author’s sinc-LLM approach to preparing prompts: six named fields, explicit rules, supplied data, output structure, versioning, and repeated checks.
Practice 1: Always Specify All 6 Bands
The format names PERSONA, CONTEXT, DATA, CONSTRAINTS, FORMAT, and TASK. Use the fields to state relevant task details explicitly, then check the response against them.
The author uses the Nyquist-Shannon sampling theorem as an analogy for task specification:
To compare prompt formats, use the same tasks, model settings, and usability criteria. Record the outputs, errors, and retries for each version.
Practice 2: Write Explicit Constraints
CONSTRAINTS states rules, limits, and exclusions to check in the output.
Good constraints are specific and easy to check:
- "Under 500 words" not "be concise"
- "No external library dependencies" not "keep it simple"
- "Must handle null inputs without crashing" not "be robust"
- "Response time under 200ms at p99" not "be fast"
- "Use only information from the provided DATA band" not "be accurate"
Practice 3: Put Real Data in the DATA Band
The DATA field identifies the supplied inputs and references. State the evidence boundary and check claims against those sources rather than relying on the model’s unstated assumptions.
Put the relevant examples, numbers, code, or quoted material in DATA. Ask for source references where the task needs them, then verify the response against the supplied material.
Practice 4: Use Specific Personas, Not Generic Ones
State the requested role and domain explicitly. For example, “Senior backend engineer specializing in PostgreSQL performance optimization” names a role and a technical focus. Check whether the response fits the requested task.
Compare role instructions using the same task and declared relevance criteria before attributing an output change to a persona.
Practice 5: State and Check the Output Format
Specify the requested structure explicitly. For example, request JSON with keys title, summary, action_items, and priority, then validate the response’s keys, value types, and content.
Use sinc-LLM to write an explicit FORMAT specification. Check the returned structure against the required fields for your task.
Practice 6: Evaluate Other Prompting Tips
Evaluate these prompting suggestions against the actual task and output checks:
- "Please" and "thank you": Zero impact on output quality. The model has no feelings to influence.
- "Think step by step" (for non-reasoning tasks): This adds tokens but does not improve output for tasks like content generation, translation, or formatting. It is only useful for math and logic with reasoning models like o3.
- Threatening the model: "Your job depends on this" or "This is very important" has zero impact. The model does not feel pressure.
- Repeating instructions: Saying the same thing three ways wastes tokens and does not help. Say it once, clearly.
- Temperature tweaking without structural changes: Lowering temperature cuts randomness but does not fix a vague prompt. A well-structured prompt at temperature 0.7 beats a vague one at temperature 0.1.
Practice 7: Test Your Prompts Systematically
Repeat a prompt with its model settings recorded and compare outputs against the same task criteria. Similar responses can still be incorrect; variation alone does not identify the cause of an error.
Keep the prompt version, settings, supplied data, expected behavior, and observed outputs together. Use the failures to identify which requirements or checks need revision.
Practice 8: Use sinc JSON for Reproducibility
Store your prompts as sinc JSON files. That makes them easy to version, share, and reuse. A prompt that only lives in a chat window will be lost.
{
"formula": "x(t) = \u03a3 x(nT) \u00b7 sinc((t - nT) / T)",
"T": "specification-axis",
"fragments": [
{"n": 0, "t": "PERSONA", "x": "Expert data scientist with 10 years ML experience"},
{"n": 1, "t": "CONTEXT", "x": "Building a recommendation engine for an e-commerce platform"},
{"n": 2, "t": "DATA", "x": "Dataset: 2M user interactions, 50K products, sparse matrix"},
{"n": 3, "t": "CONSTRAINTS", "x": "Must use collaborative filtering. Latency under 100ms. No PII in logs. Python 3.11+. Must handle cold-start users with content-based fallback"},
{"n": 4, "t": "FORMAT", "x": "Python module with type hints, docstrings, and pytest tests"},
{"n": 5, "t": "TASK", "x": "Implement the recommendation engine with train/predict/evaluate methods"}
]
}
The sinc-LLM tool builds this format for you. Use it to grow a library of reusable, structured prompts for your team.
Summary of Practices
| Practice | ||
|---|---|---|
| All 6 bands specified | ||
| CONSTRAINTS longest band | Task-specific constraints | |
| Real data in DATA band | ||
| Specific persona | ||
| Explicit format specification | Explicit format specification |
// Production AI Engineering
Build AI systems that hold up in production.
sinc-LLM designs, audits, and stabilises production AI infrastructure: from vendor evaluation and cost accountability to incident controls and MCP architecture.
See what we do →Additional context
sinc-LLM About Blog Services
© 2026 sinc-LLM. All rights reserved. DOI: 10.5281/zenodo.19152668
sinc-LLM

