Prompt Engineering

There are no magic words. Good prompting is just clear thinking made explicit: say what you want, show what good looks like, and iterate.

The practice of structuring instructions to AI systems so they produce reliable, high-quality outputs. Good prompt engineering isn’t about magic phrases or secret tricks; it’s about clarity of thought. A well-engineered prompt specifies what you want, provides enough context to constrain the output space, and iterates based on what comes back. The fundamentals are structure (how you organize the request), specificity (how precisely you describe the desired output), and iteration (treating the first attempt as a draft, not a final answer).

Here’s what that looks like in practice. A delivery lead wants the team’s retro turned into something actionable and types the obvious thing:

Summarize our sprint retro notes.

The model returns a tidy recap nobody acts on, because “summarize” handed every real decision back to the model. The engineered version says what “good” means:

You're helping a delivery lead turn a sprint retro into next-sprint action.

<notes>
[paste the raw retro notes]
</notes>

Group what you find into three themes: delivery flow, collaboration, tooling.
For each theme, name the single action most likely to move the needle next
sprint, phrased with an owner and a checkable "done" state. Drop anything
that's venting with no change attached.

Every change is pulling weight. The role fixes who the output is for. The fenced <notes> stop the model inventing input it wasn’t given. The three themes and the “owner and checkable done state” are the success criteria, not decoration. “Drop venting” is an explicit exclusion, which is where a lot of the quality comes from. None of it is a magic word; it’s the thinking you’d have to do anyway, written down where the model can see it.

We treat prompt engineering as a foundational literacy skill, not an advanced technique. Named prompt frameworks like CO-STAR and STOKE are useful scaffolding here — checklists that stop you leaving out the context, constraints, and success criteria a clear prompt needs. In our Effective AI workshop, it’s the starting point: once participants can write clear, structured prompts, they’re ready to move into more powerful techniques like reverse prompt engineering, meta-prompt engineering, and context engineering. The leverage isn’t in any single prompt; it’s in building the habit of thinking precisely about what you need before you ask.

Resources

• Effective AI — covers prompt engineering fundamentals in Session 2
• Prompt Frameworks — six named structures (CO-STAR, RISEN, STOKE, and more) for making prompts complete, with a decision matrix
• Context Engineering — the complementary skill of supplying the right background information