Create and Structure AI Prompts with Prompt Blocks

Build AI prompts from structured blocks — role, objective, context, examples, reasoning, and constraints — then expand them anywhere with a shortcut.

SprintBrain’s prompt editor breaks a prompt into up to six named blocks, each with a toggle. Enable only the blocks your task needs, fill them in, and SprintBrain assembles them into a single prompt body at expansion time. This structure makes prompts easier to read, easier to improve, and easier for the quality scorer to grade.

The prompt block editor

Every prompt is composed of six blocks in a fixed order. You can enable or disable each block independently — a simple task might only need Role, Objective, and Reasoning, while a complex one benefits from all six.

Block	Purpose	Example
Role	Define who the AI is	”You are a senior software engineer…”
Objective	State the task clearly	”Your task is to review…”
Context	Provide background	Relevant background about the situation
Examples	Show input/output pairs	Guide the model with demonstrations
Reasoning	Instruct thinking approach	”Think step by step…”
Constraints	What the model must not do	Boundaries and limitations

Each block shows a placeholder hint as you write — for example, the Examples block prompts you with “Show input/output pairs to guide the model” — so you always know what belongs in each section.

Example: a fully structured prompt

Here is what a code-review prompt looks like with all six blocks filled in:

[ROLE]
You are a senior software engineer with deep expertise in TypeScript and React.

[OBJECTIVE]
Your task is to review the code diff below and identify bugs, performance issues,
and violations of the team's coding standards.

[CONTEXT]
The codebase follows the Airbnb ESLint ruleset. The team targets Node 20 and
React 18. This is a pull request for a customer-facing feature, so correctness
is more important than brevity.

[EXAMPLES]
Input: a useState hook with a stale closure
Output: "Line 42 — the callback captures `count` from the first render.
         Use the functional updater form: setCount(prev => prev + 1)."

[REASONING]
Think step by step. First check for bugs, then performance, then style.
Double-check your result before producing the final review.

[CONSTRAINTS]
Do not rewrite the code. Flag issues only. Keep each comment under 40 words.
Do not comment on formatting — the linter handles that automatically.

Enable only the blocks your prompt actually needs. Simpler prompts often perform better than over-engineered ones — a clear Role + Objective + Reasoning frequently outperforms a six-block prompt where the extra sections are vague or redundant.

Giving a prompt its own shortcut

If you use a prompt frequently, give it a shortcut so you can expand it without opening the picker.

1. Open the prompt editor

   Click New prompt or click an existing prompt card to open the editor panel.

2. Enter a shortcut

   In the Shortcut field, type a short identifier — for example, coderev. Use only letters, numbers, hyphens, or underscores. SprintBrain sanitises the input automatically.

3. Save the prompt

   Click Create prompt (or Save changes). The shortcut is now live.

4. Use it anywhere

   In any text field, type your prompt trigger prefix followed by the shortcut and press Tab. For example, if your prefix is /, type /coderev + Tab and the full prompt expands at your cursor.

Leave the shortcut field empty to keep the prompt accessible through the """ picker only.

Dynamic fields in prompts

Prompts support the same dynamic field syntax as snippets. Use placeholders to collect values from the user at expansion time — the same inline overlay that appears for snippets appears for prompts too.

Syntax	What it does
{formtext: label="Guest name"}	Prompts for a single-line text value
{formdate: label="Check-in"}	Opens a date picker
{formmenu: default="English"; options="Spanish,French"}	Shows a dropdown
{{placeholder}}	A simple placeholder that the user fills in before inserting

Example — a support-reply prompt that asks for the customer’s name and issue at expansion time:

[ROLE]
You are a friendly customer success agent for a SaaS product.

[OBJECTIVE]
Write a reply to {{customer_name}}'s support ticket about {formtext: label="Issue summary"}.

[CONSTRAINTS]
Keep the tone warm but professional. Reply in under 100 words.

At expansion, SprintBrain pauses, shows the overlay for Issue summary, then inserts the finished prompt with all values filled in.