MEDIA PREVIEW
Overview & Result
jev-kit packages several Jev decisions around Claude Code: a PreToolUse guard for ambiguous tool calls, a check for oversized sub-agent choices, completion verification, faster file search and a bounded browser agent. Claude remains responsible for writing and planning; Jev handles small typed decisions inside the loop.
How Jev fits in the loop
- Code-only rules allow the majority of ordinary tool calls without a network request.
- Jev judges the ambiguous tool calls and checks whether a dispatched sub-agent tier is too expensive for the task.
- Claude Code executes the selected action, while the kit can send an unverified completion back for another check.
- The browser component uses Jev to choose one bounded interaction at a time and falls back to a planner for less explicit tasks.
How to reproduce
- Clone https://github.com/jonathanavis96/jev-kit and follow its installation guide.
- Run the check-only or doctor path first, then keep the guard in shadow mode before enforcing it.
- Inspect the PreToolUse guard, sub-agent tier checks, browser workflow and the repository's measured results.
- Repeat the author's small benchmarks with the same task definitions before making any cost or quality claim.
Why this build matters
This is a concrete example of Jev acting as an internal control layer for a coding agent instead of becoming another chat model.
Limitations
- The Reddit post reports small author-run measurements rather than an independent benchmark.
- The browser workflow works best when the click sequence is explicitly bounded; the author reports that open-ended navigation is not a fit.
- The guard was observed blocking nothing in 30 sessions, so the post does not demonstrate prevented incidents.
- The kit fails open when Jev is unavailable and requires a TypeSafe API key for Jev-backed checks.