Skip to content

R0020/2026-03-25/Q003/SRC03/E01

Research R0020 — Prompt Engineering Gaps
Run 2026-03-25
Query Q003
Source SRC03
Evidence SRC03-E01
Type Reported

Industry practitioners recommend imperative language and treating prompts like contracts

URL: https://medium.com/@mjgmario/prompt-engineering-basics-2026-93aba4dc32b1 (and search synthesis)

Extract

Industry guidance on imperative constraints:

Imperative language advocated: "Be strict and explicit, using strong language like 'must' to emphasize what is required." An imperative prompt tells the model "you are commissioning an output under constraints."

Contract-style prompts: "Treat prompts like contracts: a clear definition of what you want, what information the model is allowed to use, what constraints it must respect, the exact shape of the output you expect, and how the result should be checked."

Pattern: directive + constraints + format. "It works well if you force the model to follow a strong template and include the constraints you care about."

Caution on overuse: "Too many constraints can overwhelm the model, causing it to ignore some, so it's important to limit the number of constraints and focus on the most important ones."

Relevance to Hypotheses

Hypothesis Relationship Strength
H1 Supports Industry actively recommends imperative language
H2 Contradicts Explicit discussion of constraint design
H3 Supports Notes that excessive constraints can be counterproductive

Context

The "prompts as contracts" framing is notable — it elevates prompt design from casual instruction to formal specification. The caution about constraint overload is also significant: it suggests that while imperative constraints are recommended, there are diminishing returns and potential negative effects from over-constraining.

Notes

This evidence synthesizes findings from multiple industry sources encountered during the search process. The consistent theme is that imperative language is recommended but with awareness of its limitations.