> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sprintbrain.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Prompt Quality Score: Grade and Improve Your Prompts

> SprintBrain grades every prompt 0–10 across six criteria, explains each deduction, and offers one-click fixes — no prompt engineer required.

SprintBrain evaluates every prompt as you write it and shows an **Efficiency Score** directly in the editor. The score reflects how clearly and completely your prompt communicates its intent to the AI model — across the same six dimensions that structure the block editor. You see the score, understand every deduction, and fix weak areas in one click.

## How the quality score works

SprintBrain grades each prompt on a **0–10 scale** across six criteria that map directly to the prompt blocks:

<AccordionGroup>
  <Accordion title="Role (0–10 contribution)">
    Does the prompt tell the model who it is? A clear persona — "You are a senior B2B copywriter with deep SaaS experience" — anchors the model's tone and vocabulary. A missing or vague role earns a deduction.
  </Accordion>

  <Accordion title="Objective (0–10 contribution)">
    Is the task stated clearly and specifically? An objective like "Write a 150-word product announcement email for the new analytics feature" scores higher than "Write an email about the feature."
  </Accordion>

  <Accordion title="Context (0–10 contribution)">
    Does the prompt give the model enough background to avoid wrong assumptions? Context earns points for specifying the audience, the situation, and any relevant constraints the model cannot infer on its own.
  </Accordion>

  <Accordion title="Examples (0–10 contribution)">
    Does the prompt include one or more input/output demonstrations? Even a single example significantly improves model reliability. The scorer rewards both the presence and the clarity of the examples block.
  </Accordion>

  <Accordion title="Reasoning (0–10 contribution)">
    Does the prompt tell the model how to think? Instructions like "Think step by step" or "Double-check the result before producing the final answer" earn points because they improve output accuracy on complex tasks.
  </Accordion>

  <Accordion title="Constraints (0–10 contribution)">
    Does the prompt define what the model must not do? Word limits, tone restrictions, and explicit exclusions prevent common failure modes and earn points for completeness.
  </Accordion>
</AccordionGroup>

The score appears as a **circular ring** in the editor — green for 8 or above (Excellent), blue for 5–7 (Good), and amber for below 5 (Needs improvement) — along with the percentage score and a summary of how many criteria need attention.

## Reading the score

The Efficiency Score panel shows three layers of detail:

<Steps>
  <Step title="Numeric grade and label">
    The ring displays the score out of 10 (for example, **7.3/10**) alongside a label: **Excellent**, **Good**, or **Needs improvement**. This is your at-a-glance signal.
  </Step>

  <Step title="Library comparison">
    When your library has enough prompts to compare, the panel shows how this prompt ranks against the rest — for example, *"Stronger than 68% of your Coding prompts"*. Use this signal to decide which prompts to improve first and which are already pulling their weight.
  </Step>

  <Step title="Per-criterion rationale">
    Expand any criterion row to read the **rationale** — a plain-English explanation of exactly why that criterion scored the way it did — plus a concrete example of what a passing version looks like.
  </Step>
</Steps>

## One-click improvements

Next to each criterion that scored below a pass, SprintBrain shows a **suggestion label** — a small button with a short description of the recommended fix. Click it and SprintBrain applies the change immediately:

* If the fix is to **enable a block**, clicking the label turns the block on so you can fill it in.
* If the fix is to **add a specific line** (for example, "Double-check the result and flag anything you are unsure about" for a weak Reasoning block), SprintBrain appends that line to the block automatically.
* If the fix is to **set a metadata default** (for example, specifying an output type or strategy), SprintBrain updates the field for you.

You can also click any criterion row to expand it and read the full rationale and a concrete example before deciding whether to apply the suggestion.

## Team baseline

When you share a prompt folder with your team, everyone draws from the same prompt library and the same quality baseline. High-scoring prompts raise the output quality across the whole team without requiring a dedicated prompt engineer — your teammates benefit from the work you put into structuring and improving prompts the first time.

The Efficiency Score is visible on each **prompt card** in the dashboard, so anyone browsing the library can immediately see which prompts are well-structured before using them.

<Tip>
  Review low-scoring prompts before sharing a folder with your team. A prompt that scores below 5 often has a missing role or an unclear objective — the two deductions that cost the most points and are the fastest to fix.
</Tip>

<Note>
  The quality score is advisory. It helps guide structural improvements, but a well-tested, battle-hardened prompt with a score of 6 may outperform a freshly written 9 on your specific task. Use the score as a starting signal, not a final verdict.
</Note>