Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

orca-jev-advisor

Orca Lab plugin where Jev judges each agent command, passing routine ones, refusing dangerous ones and giving reasons for risky ones.

Projectbeta
View on GitHub ↗

Overview

Jev Advisor is an Orca Lab plugin that judges what agents are about to do before they do it. A fast Jev judgment sits in front of every command an agent issues: routine commands like npm test or rm -rf node_modules pass silently, dangerous ones like terraform apply or git push --force are refused outright, and for other commands Jev sees as risky, the coding model receives a concrete reason instead of interrupting the user. A person is asked only when team policy says a human must decide.

The plugin also has a Memory module that, in Active mode, saves what is worth remembering to Engram or to Claude Code's auto-memory folder, with Haiku used to resolve memory conflicts and file-based fallbacks when Engram is unavailable. The README emphasises that judgments take a few hundred milliseconds for a fraction of a cent.

What the decision model decides

TypeQuestionAnswers
ChoiceWhat should happen with this agent command?(inferred)allow · refuse · advise · ask human
Yes / noIs this worth remembering?(inferred)yes · no

How it works

  1. Agent proposes a shell command.
  2. Jev judges its risk.
  3. Allow, refuse, or return a reason to the model; ask a human only per policy.
  4. Optionally save useful memories after turns.

Strengths

  • Fewer interruptions for routine commands.
  • Dangerous commands blocked outright.
  • Fast and cheap judgments.

Caveats

  • Orca Lab-specific.
  • Memory module adds Haiku calls and Engram dependency.

Getting started

Install via Orca Lab Settings > Plugins and add your TypeSafe key.

At a glance

Type
Project
Status
beta
Language
TypeScript
Runtime
Orca Lab plugin
Has code
yes

Use cases and integrations

How curated lists describe it

Frequently asked questions

What does Jev decide?

Whether each agent command is routine, dangerous or risky enough to warrant a reason.

When is a human asked?

Only when your team policy requires a human decision.

What does the Memory module do?

Saves useful memories to Engram or Claude Code's auto-memory folder.

Related projects

  • claude-jev-plugin

    Claude Code plugin adding Jev judgments to Bash, Write and Edit calls and their outputs; a semantic guardrail, not a sandbox.

    Pluginbeta
  • interlinked-cli

    Local guard layer for coding agents that blocks dangerous tool calls by deterministic rules, with optional Jev judgments.

    CLIbetaTypeScript
  • jev-axi

    Agent-ergonomic Jev CLI and PreToolUse hook that scores shell commands for hazards and screens fetched text for injection.

    PluginbetaTypeScript
  • jev-firewall

    PreToolUse firewall for Claude Code and Codex: shell-aware rules first, then Jev returns allow, ask or block, failing closed.

    HookbetaTypeScript
  • jev-guard

    Safety gate plugin for Claude Code, Codex CLI and Antigravity: local sensitive-file checks, then Jev judges blast radius.

    PluginbetaTypeScript
  • jev-guard (leepokai)

    Security hook giving Codex, Copilot, Gemini, Cursor, pi, OpenCode and ACP an auto-mode-like guard via three Jev questions.

    LibrarybetaTypeScript