A Claude Code and Codex skill that fires before code is written, splits a feature into atomic steps and assigns each one Jev, an LLM, or plain code.
User repository on GitHub
Makes Claude Code and Codex ask whether a step needs a generative LLM at all — or whether it belongs on Jev, TypeSafe's System One model. Architecture arbitrage before the code is written.
Repo stats from the GitHub API, cached Oct 2026.
This is one of several Jev listings for the same project. The canonical entry is Jev Architect — karanb192.
A skill that makes Claude Code and Codex ask one question before they write your code: does this step really need a generative LLM?
Quoted from the project's README. Read the full README.
Watch out
The gate is a set of criteria applied by a language model, not a measurement, so it will sometimes stay silent where Jev would have helped and sometimes propose it where integration cost is not worth it. Volume and latency numbers usually are not in the code, so it is told to write to confirm rather than invent a row. Order-of-magnitude claims come from vendor documentation rather than a benchmark of your workload, and it proposes a migration rather than replaying real history to compare decisions. Jev is text-only and constrained to the options you enumerate. It adds no runtime dependency, is not an API reference, and is complementary to TypeSafe's own skill rather than a replacement. Copy reviewed against the repository README on 2026-10-04, not from running it.
skeptrunedev/jev-recruiterPython
win4r/jev-skill-suggesterPython
PyModel/jev-judge-mcpPython
rhighs/jev-codeTypeScript
shaharia-lab/jev-cliRustPosts from builders shipping with Jev right now.
