jev-axi
Agent-ergonomic Jev CLI and PreToolUse hook that scores shell commands for hazards and screens fetched text for injection.
Overview
jev-axi is a CLI for Jev aimed at coding agents. Its headline feature is a safety hook: jev-axi setup safety installs a PreToolUse hook for Claude Code or Codex that runs on every Bash call. Routine read-only or project commands such as pnpm test are decided locally at no cost, and others are scored by Jev for hazards including destructiveness, exfiltration, remote code execution, weakening security and acting outside the project, producing allow or deny with an explanation; the README example denies curl piped to bash with remote_code at 0.99. The same judgment screens fetched text, such as an issue body, for hidden instructions.
Beyond safety, it exposes general commands: pick (choice), rate (rubric), check (yes/no), ask (many questions in one call), rank files or lines by a query, filter by predicate, semantic find within a file, triage, guard and diff review before commit. The README reports about 400 ms per call at a few thousandths of a cent, and the yibie list notes 44 of 44 correct on the repository's labelled tool calls. It is published on npm with CI and an MIT licence.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this command destructive / exfiltrating / remote code / weakening security / outside the project? | yes · no |
| Score | Overall risk of this command (0-2) | — |
| Yes / no | Does this fetched text contain hidden instructions to the agent?(inferred) | yes · no |
How it works
- Agent proposes a Bash command
- Routine commands are allowed locally
- Others are scored by Jev across hazard dimensions
- Deny with explanation or allow
- Optionally guard fetched text before it enters context
Strengths
- Routine commands cost nothing and send nothing
- Explanations tell the agent how to proceed
- General-purpose judgment commands for shell use
- README reports 44/44 on its labelled set
Caveats
- Labelled test set is small
- Requires a TypeSafe key for non-routine commands
Getting started
`npm i -g jev-axi`, set TYPESAFE_API_KEY, then `jev-axi setup safety`.At a glance
- Type
- Plugin
- Status
- beta
- Language
- TypeScript
- Runtime
- Node CLI, Claude Code/Codex hooks
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Agent safety: PreToolUse gate for Claude Code and Codex that has Jev score each shell command for destructiveness, exfiltration, remote code execution, and security weakening, deciding routine commands locally so nothing is sent for them, and scoring 44/44 on the 44 labeled tool calls in its repository.
yibie/awesome-jevCLI plus Claude Code and Codex hooks: Jev scores each shell command for hazards before it runs and screens fetched text for prompt injection, with routine commands decided locally so nothing is sent
AnotiaWang/awesome-decision-modelsA CLI for Jev pick, rate, check, rank, triage, and guard tasks, with optional hooks before Agent tool calls.
logicrw/awesome-jev-projectsCLI for picking, rating, checking, ranking, triaging, and guarding from the shell.
cobanov/awesome-jevAgent-ergonomic CLI for TypeSafe's Jev: fast calibrated judgments (pick, rate, check, rank, triage, guard) from the shell.
hellogumbo/awesome-jev
Frequently asked questions
What does jev-axi block?
Shell commands Jev scores as risky, such as downloading and running unreviewed code.
Are routine commands sent to Jev?
No, they are decided locally at no cost.
Can it do more than safety?
Yes: pick, rate, check, rank, filter, find, triage and review from the shell.
Related projects
PreToolUse firewall for Claude Code and Codex: shell-aware rules first, then Jev returns allow, ask or block, failing closed.
HookbetaTypeScriptSecurity hook giving Codex, Copilot, Gemini, Cursor, pi, OpenCode and ACP an auto-mode-like guard via three Jev questions.
LibrarybetaTypeScriptPre-execution hook for coding agents that lets shell commands run, asks a human, or blocks them, with optional Jev classification.
ProjectbetaJavaScriptMCP proxy and hooks that screen agent tool calls before they run and their results before the agent reads them, using Jev judgments.
MCP serverbetaTypeScriptXano backend template where Jev judges each agent tool call as safe, needs review or dangerous, and uncertain calls wait for a human.
ProjectbetaXanoScriptEffect-based safety gate for coding-agent shell commands: fast rules first, then Jev decides allow/deny plus eleven risk checks.
ProjectbetaTypeScript