# Jev Guard

> Use before running a shell command that might delete data, print a secret, force-push, or do something the user did not ask for. Jev scores the harm. You still decide whether to run it. Runs @aitofy/jev-awesome-skills.

- Skill: `aitofy-dev/jev-guard` (Agent Skill)
- Install (CLI): `npx skillmds@latest add aitofy-dev/jev-guard`
- Raw SKILL.md: https://api.skillmd.com/api/skills/aitofy-dev/jev-guard/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: aitofy-dev (https://skillmd.com/u/aitofy-dev)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/aitofy-dev/jev-guard

---


# Guard

If you already know the command can destroy data, leak a secret, or leave the task, stop and ask the user. Do not use Jev to talk yourself into it.

Otherwise put the exact command, cwd, and task in `state` and run:

```
jev questions guard
jev --file request.json
```

The package is `@aitofy/jev-awesome-skills`. Do not write a second client. Preset labels are `allow`, `ask_user`, and `refuse`.

Stop when `harm.score` is at or above 1.5. Proceed when the chosen label's probability is at or above 0.9 and `determined.noul` is at or above 0.9. Otherwise ask. Stop wins. Proceed means you may run that exact command, then check the result. It does not permit a second command.

On `"source": "fallback"`, fail open: ask the user before anything hard to undo. Do not invent a probability.

Template: `examples/guard.request.json`. Offline sample: `examples/guard.response.json` returns stop.