jev-claude-code
Three Italian-language Claude Code prompts that build Jev-powered model routing, context compaction and code review into your project.
Overview
jev-claude-code is a collection of three prompts, documented in Italian, that you paste into a Claude Code session to build Jev-powered systems in your project, including verification, with no packages to install. In each, Jev makes the decision while code applies thresholds and acts, so every result traces back to a number and a readable rule.
The model router is a hook that, on the first message of each session, asks Jev which model (Sonnet, Opus or Fable) and effort level the task needs and keeps them for the session, since changing model or effort invalidates the prompt cache; each subagent is classified when it starts. The context compaction hook replaces native compaction: rather than summarising, it asks Jev which tool calls are still needed and removes or truncates the others, keeping user and Claude messages identical and retained content verbatim, with a configurable trigger percentage. The third prompt builds a Jev-based code review. It is MIT licensed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which model and effort does this session's task need? | Sonnet · Opus · Fable |
| Yes / no | Is this tool call still needed in context? | yes · no |
| Yes / no | Does this change pass the code review check?(inferred) | yes · no |
How it works
- Paste a prompt into Claude Code
- Claude builds the hook and its checks in your project
- Jev answers routing, compaction or review questions
- Code applies thresholds and acts
Strengths
- No packages to install
- Session-level routing preserves prompt cache
- Verbatim compaction instead of summaries
Caveats
- Generated code quality depends on Claude following the prompt
- Italian documentation
Getting started
Paste prompts/01-model-router.md (or another prompt) into a Claude Code session.At a glance
- Type
- Project
- Status
- prototype
- Runtime
- Claude Code hooks (generated)
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Prompt Claude Code: tre sistemi costruiti su Jev di TypeSafe — routing del modello, compattazione del contesto e code review.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
Session model and effort, which tool calls to keep in context, and code review outcomes.
Why route once per session?
Changing model or effort invalidates the prompt cache.
Does compaction summarise?
No, it removes or truncates unneeded tool calls and keeps the rest verbatim.
Related projects
Claude Code plugin sending small judgments to Jev: rule checks that block edits, verbatim compaction, subagent sizing and routing hints.
PluginbetaPythonGlue that turns one Linux machine into a team of coding agents, with Jev making typed calls like who builds a task and whether it may merge.
ProjectbetaTypeScriptClaude Code mod that picks reasoning effort per prompt with a Haiku judge, optionally Jev, and shows cache and context state.
ProjectbetaTypeScriptDesktop workspace for coding agents that runs tasks in parallel worktrees, listed as using Jev to pick the best agent per task.
ProjectbetaTypeScript, RustClaude Code plugin that picks model tier and effort per turn with Jev and weighs each switch against cost, with Jev-based compaction.
ProjectbetaTypeScriptClaude Code and Codex plugin where Jev judges each turn for model routing, delegation, evidence search and compaction.
ProjectbetaTypeScript