Practice Lab is the home of Confidence — the third pillar. Frontier AI fails invisibly: confidently, plausibly, in ways only a domain expert spots. The Lab is the workspace where you catch it before it ships.
Click Practice Lab in the sidebar, or use the "Open in Lab" button from the Prompt Composer.
What you'll see
Prompt panel
Your prompt as Keypra parses it — broken into the underlying blocks (role, context, task, constraints, output).
Score & findings
A 0–100 quality score plus specific, actionable findings: missing context, ambiguous instructions, conflicting constraints.
Stress Test
On-demand AI review that surfaces tensions and blind spots between your blocks before you send anything live.
Refinement loop
Apply suggested fixes one at a time and see the score update. Save the improved version straight back to your Library.
How to use it
Send a prompt to the Lab
From the Composer, click "Open in Lab". Or paste any prompt directly into the Lab's input.
Read the findings, not just the score
A high score with one critical finding still needs work. The findings tell you why.
Run a Stress Test
Especially before client-facing or high-stakes work. The test names tensions you might miss.
Refine, then save
When the prompt holds up, save it back to your Library so future-you starts from a stronger draft.
Composer ↔ Lab bridge
The Composer and the Lab share state through your session, so you can bounce between them without copy-pasting. Build in the Composer, test in the Lab, refine, then build again.
You don't have to use the Lab on every prompt. Reach for it when stakes are high — legal language, financial figures, anything client-facing — and it pays back the few extra seconds many times over.