Agent Guard

A local safety-check server that lets an assistant screen planned actions for malware, destructive commands, and secret leaks before acting.

What you can do

Agent Guard provides pre-action safety checks an agent can run before it executes something. According to its description, it covers three areas: package malware, destructive commands, and secret leaks. It runs as a local process over stdio, so the checks happen on the user's own machine rather than through a remote service.

When to use it

Worth adding when an assistant is allowed to install packages, run shell commands, or handle credentials and you want a guardrail step before those actions take effect. It suits setups where the agent's actions carry real risk of damage or exposure.

Requirements

Runs as a local process on the user's machine over stdio, installed from the PyPI package agent-tripwire, so a Python runtime and package installation are needed. No authentication, environment variables, or headers are declared. Desktop only; not available as a web executable.