Overview & result
An agent skill for writing and improving programs that call Jev. The README says it covers question design (Choice, Score, Noul), state structure, composing answers in code, confidence thresholds, and diagnosing questions that answer wrong or with low confidence. It installs as a Claude Code plugin, through the skills CLI, or by copying the skill folder.
From the original post
A skill for writing and improving programs that call Jev, TypeSafe's System One model
How Jev fits in the loop
- Install the skill as a Claude Code plugin, through the skills CLI, or by copying the folder into your skills directory.
- Claude loads it on its own when a task involves Jev or TypeSafe questions, or you invoke it with /jev.
- The skill guides question design (Choice, Score, Noul), state structure, composing answers in code and setting confidence thresholds.
- It also covers diagnosing questions that answer wrong or with low confidence.
How to reproduce
- Inspect the original showcase and source material at https://github.com/dbreunig/building-with-jev-skill
- Verify the bounded prompt and input candidate schema configured for Jev
- Benchmark decision latency and classification accuracy against baseline models
Why this build matters
It is guidance for the agent writing Jev code rather than a runtime tool, drawn from the TypeSafe docs.
Limitations
- It targets jev-1.13, so later model versions may differ.
- The repository is the skill only; it does not call Jev itself.
- The README gives no measurements of how much it improves the code an agent writes.