CommunityEvaluation & safety
Jev Guard
leepokai/jev-guard
A coding-agent safeguard that uses session context to assess tool-call and prompt-injection risks.
What problem does it solve?
A coding-agent safeguard that uses session context to assess tool-call and prompt-injection risks.
What public sources establish
Assesses tool actions as allow, ask or deny.
Checks risk signals in tool results and skills.
How to read and try it
Consider test conditions, versions and failure cases. Vendor or author reports are not independently verified conclusions from this directory.
Check before integrating
01
Read the original source
Review the author notes or official docs for current status, dependencies and limitations.
02
Validate the smallest use case
Start with a small test before connecting a real workflow. This directory has not run the project for you.
03
Keep fallbacks and source records
Record the version, configuration and sources used so later updates remain traceable.
Sources & usage boundaries
A safeguard is not a security guarantee. Combine deterministic permissions with independent tests. No security audit was performed here.
Author repository / READMEhttps://github.com/leepokai/jev-guard
Source review: 2026-09-19 · Scope: public pages, author descriptions and resource positioning. No independent execution, performance retesting or security audit.