omp-jevens-classifier
OMP plugin that sends unreviewed bash commands to Jev and turns UNSAFE or UNSURE verdicts into Run-or-Deny prompts.
Overview
omp-jevens-classifier adds Jev-judged permission checks to OMP's bash tool and to eval payloads that spawn processes. In yolo mode OMP auto-approves every bash command, including destructive ones, and prompt rules for critical shapes never fire because critical-pattern matches outrank them. This plugin registers a tool_call handler in front of the native bash tool and sends commands that would otherwise run unseen to Jev.
Calls are decided in order: critical patterns always prompt without a Jev call; commands with caller-supplied env prompt; deny, prompt and narrow allow rules are left to the native gate; blanket allows or unmatched commands are classified, with SAFE passing through and UNSAFE or UNSURE raising a plugin prompt. Session grants for a directory outrank classification. The handler only adds friction, never bypassing the native gate or executing anything. A system document describes its layered design.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Is this bash command safe to run without asking? | SAFE · UNSAFE · UNSURE |
How it works
- Intercept bash and process-spawning eval calls.
- Apply critical patterns, env checks and existing rules first.
- Classify remaining commands with Jev.
- SAFE runs; UNSAFE or UNSURE prompts Run-or-Deny.
Strengths
- Closes the yolo-mode gap for destructive commands.
- Never weakens existing rules or bypasses the native gate.
- Routine commands stay silent.
Caveats
- Sends commands to TypeSafe.
- Classification errors could still let a risky command run if judged SAFE.
Getting started
Install the plugin into OMP and set your TypeSafe key.At a glance
- Type
- Project
- Status
- beta
- Language
- TypeScript
- Runtime
- Oh My Pi plugin
- Has code
- yes
Use cases and integrations
How curated lists describe it
Jev-powered model-judged permission gate for OMP (TypeSafe System One).
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
Whether an otherwise unreviewed bash command is SAFE, UNSAFE or UNSURE.
Can it auto-approve dangerous commands?
It only adds friction; critical patterns always prompt without consulting Jev.
Does it change my rules?
No, deny, prompt and narrow allow rules are left to the native gate.
Related projects
Local guard layer for coding agents that blocks dangerous tool calls by deterministic rules, with optional Jev judgments.
CLIbetaTypeScriptSecurity hook giving Codex, Copilot, Gemini, Cursor, pi, OpenCode and ACP an auto-mode-like guard via three Jev questions.
LibrarybetaTypeScriptPi extension that turns what you say into structured runtime constraints and checks every side-effecting tool call against them.
PluginbetaTypeScriptPi extension using Jev as a tool-call gate, a bash output judge and a jev_ask tool, in shadow mode by default and failing open.
PluginbetaTypeScriptAuto mode for Pi: deterministic policy first, then Jev judges escalated bash/write/edit calls, failing closed when undecidable.
ProjectbetaTypeScriptPi extension that auto-approves routine Bash commands via a System One model and escalates flagged ones, including obfuscated, to a human.
PluginbetaTypeScript