promptstyles.org
Taxonomy v0.1.0

Prompt styles

A shared vocabulary for the ways engineers direct AI when they write software. Each style has a definition, an example prompt, what it reveals about the person using it, when it fits, what to watch out for, and where the term came from.

Style map

DeclarativeImperativeVibe codingSpec-driven developmentTest-driven promptingExample-based (few-shot)Iterative / conversationalContext engineering
Horizontal: how much the prompt dictates the implementation. Vertical: how much the output is checked against an explicit contract (tests, a spec, review). Styles are a spectrum, not exclusive boxes. Strong engineers combine them and switch on purpose as risk and complexity change. Positions are approximate and open to debate: propose a change with a pull request.

One feature, eight ways

Every style shows a prompt for the same feature so you can compare them side by side: an image carousel for a home page that shows 5 images from the database, advances every 5 seconds, lazy-loads images and flashes twice when switching.

  1. 01

    Declarative

    Describe what the software must do and let the AI decide how.

    Read more →
  2. 02

    Imperative

    Dictate how the code should work: data structures, algorithm, control flow.

    Read more →
  3. 03

    Vibe coding

    Describe loosely, accept what comes back, barely read the code.

    Read more →
  4. 04

    Spec-driven development

    Write a specification first. The AI implements from it.

    Read more →
  5. 05

    Test-driven prompting

    Write the tests first and ask the AI to make them pass.

    Read more →
  6. 06

    Example-based (few-shot)

    Point to an existing pattern and ask for the same thing, adapted.

    Read more →
  7. 07

    Iterative / conversational

    Short cycles of ask, review and correct.

    Read more →
  8. 08

    Context engineering

    Engineer what the model sees (files, docs, rules, tools) more than the wording.

    Read more →

Taxonomy as JSON