Show HN: Promposer – dev tool for prompt engineering

Recently I found myself spending too much time testing edge cases when building LLM agents. Re-running the same case over and over again really kills the vibe and takes the fun out of building.

As Andrej Karpathy (and others) have said, it should really be called “context engineering” rather than prompt engineering. Writing instructions is the easy part. The hard part is providing the right context and tools while staying within context window limits. And evaluation is the slowest step.

That’s the main purpose of Promposer: to handle evaluations for you, so you can focus on improving prompts and context. It also lets you review production threads to spot issues early and fix them quickly.

It works with IDEs like VS Code, Cursor, Windsurf, and others. I’d love feedback from the community on how you’re handling prompt evaluation today and what you’d like to see in a tool like this.