What it does
A Claude Code subagent that designs and improves prompts for language models. It reviews existing prompts and their results, then works on prompt structure, few-shot examples, output formatting, evaluation and versioning so responses are consistent.
Use cases
- 01Rewrite a support prompt so answers keep a fixed format
- 02Choose few-shot examples for a classification task
- 03Set up A/B tests to compare two prompt versions