jevify_
Kevin-Zhouu / jevifyAgent skill

Get your agents running
on Jev today

Illustrative shell demo: run Jevify, see the plan, approve a branch, review the conversion

Illustrative demo · sample files, counts and test results. About this demo ↗

One command. See where Jev fits in your repo.
Approve the plan. Keep your original LLM as fallback.

Codex · Claude Code (beta) / Evaluation status ↓
$npx skills add Kevin-Zhouu/jevify

Run in your project. Choose your coding agent in the installer.

LLM vs Jev for decisions

Raw data

12 questions → 136 lines of JSON.

Jev 0.44sClaude Haiku 4.5 5.11sCategory agreement 12/12

A single synthetic example via OpenRouter, replayed at actual speed. Probability and confidence values differ. Not a general benchmark.

How it works

No key needed to audit
Claude Code

/jevify

In Codex: $jevify. The current repo is the default. Add a file or folder to narrow the audit. No setup questions before the plan.

01 / Audit
Find decision calls and flag unsuitable ones.
02 / Approve
Choose what changes, and where: branch, clone, or report.
03 / Convert
Keep interfaces intact and add the approved fallback.
04 / Validate
Run tests. Compare real answers, latency, and cost with a key.

Just exploring? Run /jevify and choose “No, just save this report.”

Built with Jev

Source code

Minecraft: taking down the Ender Dragon.

Minecraft × Jev + AstraWatch the Ender Dragon demo Original video by Ronak Malde ↗

Community showcase, not a Jevify conversion. Pre-surveyed seed, Peaceful mode.

Bring your key

Use TYPESAFE_API_KEY or OPENROUTER_API_KEY through your environment or the hidden-input launcher. Keep your original provider key for fallback. No Jev key is needed for an audit; live validation requires one.