Pi Jev Guard
Pi guard plugin with timing-based rules, Jev risk checks on tool calls, output redaction and reminders for repeated failures.
Overview
Pi Jev Guard (zszz3) is an experimental Pi coding-agent plugin, documented in Chinese, that lets you configure rules by timing interactively with /jevguard add: choose when to check, whether to use a local match or a Jev judgment, and what to do on a hit. Built-in behaviours confirm recursive deletes, discarded Git changes, force pushes and credential-file access; for other shell, write, edit and custom tool calls, one Jev check asks four questions at once about destructiveness, unauthorised data exfiltration, task drift and rule conflicts.
It also reminds the agent about apparent AGENTS.md violations, flags the third identical error from the same arguments, reminds about running verification after source or config writes, and redacts known credential patterns in tool output locally, optionally using Jev to judge residual credentials. Without a key it makes no network requests and only local rules run. Keys are set with /jevguard login, validated with a fixed test request and stored with 0600 permissions. An observe mode downgrades blocks to notices.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this tool call destructive? | yes · no |
| Yes / no | Does it exfiltrate data without authorisation? | yes · no |
| Yes / no | Does it drift from the task or conflict with a rule? | yes · no |
| Yes / no | Does this output still contain a credential? | yes · no |
How it works
- Configure rules by timing and action with /jevguard add.
- Local rules run first; Jev answers four questions for other calls.
- Confirm, block or remind based on the rule action.
- Redact credentials in output and track repeated failures.
Strengths
- Interactive rule creation without code changes.
- Works locally without a key.
- Observe mode for trial runs.
Caveats
- Experimental; Chinese documentation.
- Credential uploads are not specially blocked by default.
Getting started
pi install https://github.com/zszz3/Pi-Jev-Guide, then /reload, /jevguard login and /jevguard add.At a glance
- Type
- Plugin
- Status
- prototype
- Language
- TypeScript
- Runtime
- Pi plugin
- Has code
- yes
Use cases and integrations
How curated lists describe it
Pi coding-agent plugin with rules configured by timing, plus risk checks, output redaction, and reminders on repeated failures. Chinese documentation.
MrJev/awesome-jev
Frequently asked questions
Does it work without a TypeSafe key?
Yes, only local rules, redaction and evidence tracking run.
How are rules added?
Interactively with /jevguard add.
Where is the key stored?
In ~/.pi/agent/jev-guard/auth.json with 0600 permissions.
Related projects
Pi 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.
PluginbetaTypeScriptPython/TypeScript/Go chatbot safety library checking user input and bot replies with Jev and returning one actionable verdict per message.
LibraryproductionPythonSafety gate plugin for Claude Code, Codex CLI and Antigravity: local sensitive-file checks, then Jev judges blast radius.
PluginbetaTypeScriptClaude Code mod that answers auto mode's tool-call permission question with eight Jev yes/no checks behind a shell-aware blocklist.
ProjectbetaTypeScriptRing Zero: kernel-level runtime security for AI agents, with Jev as an optional check that can only raise severity, never permit.
ServicebetaClaude Code hook that blocks known key formats locally and sends unknown high-entropy strings to Jev only in masked form.
HookbetaPython