Sandbox
An offline practice surface that scores the quality of a prompt or code snippet, with a worked quality ladder. No AI model is involved.
What it is
The sandbox is a client-side prompt and code grader. A learner pastes a prompt or code snippet, and a deterministic, in-browser evaluator scores it across four dimensions (clarity, specificity, structure, depth) and returns a 0 to 100 score, a verdict, and specific feedback tips.
It does not call any AI model: the evaluation runs entirely in the browser as a local heuristic. A separate 'Watch quality scale' panel shows the same request at three levels (Unclear, Clear, Elite) with per-dimension scores, so learners can see exactly what separates a weak ask from a strong one.
Use
Open Sandbox from the nav, choose Prompt Lab or Code Lab, paste your text, and click Evaluate Input. Practice is untracked: nothing here counts toward XP or completion.
Setup
No setup required. The sandbox is available from the top nav to any signed-in user.
Security
- Evaluation is fully client-side: the input never leaves the learner's browser and no model request is made.
- Sandbox is an authenticated surface (requires sign-in).