A tool that tests and improves AI agent prompts automatically
is a beta tool for reducing guesswork when improving prompts for . Small wording changes can look better after a few sample answers, but they may quietly make other tasks worse. The tool treats prompts like software that needs .
Users write a plain-language rubric with weights for what matters, such as accuracy, tone, and whether the answer resolves the task. The engine runs the prompt across an eval dataset, scores the outputs, rewrites the instructions, and scores them again until it finds a stronger version. In the example shown, an output score rose from 0.62 to 0.94.
The beta is limited and aimed at people actively building and testing , with access codes sent by request.
Key points
- automatically evaluates and improves prompts for .
- Users define a weighted rubric in plain language, including things like accuracy, tone, and task .
- The engine tests the prompt on an eval dataset, scores it, rewrites it, and tests again.
- The shared example claims a score increase from 0.62 to 0.94.
- Access is through a limited beta for people building and testing .