What it does
Only you can pause it, so the agent cannot dodge a block it just hit. Pause and resume are flag files, because each hook runs as a fresh process.
Agent tooling
JevGuard
A Claude Code and Codex plugin that checks each tool call and final answer against the project's own rules with a System One model.
Bookmark: JevGuard
Keep this in your collection.
Leave a noteWhat would you try with this?
: JevGuard
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
A developer can use this tool to add strict rules to an AI coding assistant. They write specific limits in a text file, like forbidding leftover test messages. The plugin checks every proposed change and blocks the assistant from writing code that breaks those rules.
The plugin requires an access key that connects it to the outside TypeSafe service to evaluate the rules. The AI assistant cannot pause this check to bypass a block. Only a human user can temporarily turn the safeguard off while working on a project.