Skip to content
English
Back to platform
  • There are no suggestions because the search field is empty.

What Makes a Strong Prompt

A simple framework for writing prompts that get you closer to the output you need—on the first try.

Here's the good news: you don't need to write the perfect prompt. Give Sidekick a task and a geography, and it will ask clarifying questions to fill in the gaps. But we've also learned that the more context you give up front, the better your output matches what you actually need. We think of the best prompts as having three layers.

The Framework: Three Layers

Every strong prompt has up to three layers:

Task + Context + Constraints

Layer Question It Answers
Task What are you making, and where?
Context Why, for whom, and what should Sidekick know?
Constraints How should it look, what should it include, what should it emphasize?

Layer 1: Task (Required)

What are you making, and where? This is the only required part of your prompt.

Name the deliverable type and the geography.

For example, Task:

Create a housing needs assessment for Ada County, Idaho.

That's it. That's a valid prompt. Sidekick will ask follow-ups—who's the audience, what insights matter, what's the goal—and build your deliverable from there. If you're short on time or just exploring, start here. This works.

Layer 2: Context

Why are you creating this? What should Sidekick know?

Context answers questions like:

  • Why are you creating this?
  • Who is your audience?
  • What's the situation or challenge?
  • What points need to come through?
  • What outcome are you hoping for?

You don't need answers to all of these. Just add what you know.

For example, Task + Context:

Create a housing needs assessment for Ada County, Idaho.

The audience is state legislators. The goal is to build the case for workforce housing funding. We need to show affordability gaps for middle-income households—the people who earn too much for assistance but can't afford market-rate housing.

Now Sidekick understands the stakes, the audience, and the angle you need. This is the layer that makes your deliverable feel like yours, not just a generic report.

Layer 3: Constraints

How should this output look, feel, and be structured? This is the expert layer. You're telling Sidekick exactly how to shape the output.

Constraints answer questions like:

  • How should it look and feel?
  • What should it include or exclude?
  • What should it emphasize or de-emphasize?
  • Any data preferences?
  • Any tone or framing requirements?

For example, Task + Context + Constraints:

Create a housing needs assessment for Ada County, Idaho.

The audience is state legislators. The goal is to build the case for workforce housing funding. We need to show affordability gaps for middle-income households—the people who earn too much for assistance but can't afford market-rate housing.

Use one-year ACS data where available. Include plenty of maps—our legislators respond well to visuals. Use asset-based language throughout. Keep the executive summary to one page.

Now Sidekick knows your data preferences, format preferences, tone, and constraints. This level of specificity means less editing on your end. The first draft already reflects how you work and what you need.

All Three Work

Level What It Gets You
Good (Task only) A solid starting point. Sidekick asks questions to fill the gaps.

Better (Task + Context)

 A deliverable shaped to your audience, goals, and angle.
Best (Task + Context + Constraints) A first draft that's nearly ready to refine.

You don't have to start at "best." But now you know how to get there when you're ready.