42-evey
evey-email-guard
Dual-layer prompt injection screening for Hermes agents: 20+ regex patterns plus a local AI classifier.
Repository last updated Jul 18, 2026
Overview
Evey Email Guard is a plugin for the Hermes Agent ecosystem that protects autonomous agents from prompt injection attacks. It combines two complementary detection layers: a rule-based engine with over 20 regex patterns and a local AI classifier that runs on-device without external API calls.
The plugin intercepts incoming messages and evaluates them for malicious content designed to hijack agent behavior. The regex layer catches known attack patterns instantly, while the AI classifier provides adaptive detection for novel or obfuscated injections. This dual approach balances speed and accuracy.
Built for developers deploying autonomous agents in production, Evey Email Guard integrates directly into the Hermes Agent plugin system. It requires no cloud dependencies, making it suitable for privacy-sensitive or air-gapped environments.
Highlights
- Dual-layer detection: regex + local AI classifier
- 20+ regex patterns for known injection attacks
- On-device AI classifier, no external API calls
- Designed for Hermes Agent autonomous agents
- Privacy-friendly, works in air-gapped environments
Features
Regex Pattern Matching
Over 20 curated regex patterns detect common prompt injection techniques instantly.
Local AI Classifier
A lightweight on-device model identifies novel or obfuscated injection attempts without cloud dependencies.
Hermes Agent Integration
Plugs directly into the Hermes Agent plugin system for seamless protection of autonomous agents.
Installation
Commands are reproduced exactly as published by the project maintainers. Always check the upstream repository for the current instructions.
git clone https://github.com/42-evey/hermes-plugins.git
cp -r hermes-plugins/evey-* ~/.hermes/plugins/
cp hermes-plugins/evey_utils.py ~/.hermes/plugins/
# Restart hermes-agentRequirements
- Python 3.8+
- Hermes Agent installed
How it's built
The plugin is written in Python and follows the Hermes Agent plugin specification. It consists of two main files: evey-email-guard/__init__.py contains the detection logic, and evey-email-guard/plugin.yaml declares the plugin metadata and configuration.
The regex layer uses a compiled set of patterns stored in the plugin, while the AI classifier is a small model loaded at startup. Both layers run synchronously on incoming messages, returning a risk score or classification. The plugin does not require any external services or network access.
Source
This summary was written for Neura Market from the project's own repository. Install commands, identifiers and licence details are reproduced unchanged. The 42-evey/evey-email-guard repository is the authoritative source and the place to file issues or contribute.
Category: Security & Privacy · Licensed under MIT