Introducing Sage: A Security Layer for AI Agents

AI agents are increasingly being used on developer workstations, executing shell commands, fetching URLs, and writing files with little inspection. However, this lack of oversight poses significant security risks. Therefore, the open-source project Sage has been developed to insert an interception layer between AI agents and these operations, checking each action before it proceeds.

What is Sage and How Does it Work?

Sage works through hook systems native to the agent platforms it supports, including Claude Code, Cursor/VS Code, and OpenClaw. Meanwhile, it intercepts tool calls, such as Bash commands, URL fetches, and file writes, and passes them through several detection layers. For example, URL reputation checking runs cloud-based malware, phishing, and scam detection.

Additionally, local heuristics use YAML-based threat definitions for dangerous patterns, and package supply-chain checks cover registry existence, file reputation, and age analysis for npm and PyPI packages. Furthermore, plugin scanning runs at session start and checks other installed plugins for threats.

Threat Data and Research

The release of Sage connects to research from Gen Threat Labs, which found over 18,000 OpenClaw instances currently exposed to the internet and open for attack. Moreover, nearly 15% of observed skills contained malicious instructions. Siggi Stefnisson, Cyber Safety CTO of Gen, described the situation as a shift in which security failures are no longer just one bad click, but trusted AI assistants quietly turning into persistent insider threats.

Finally, to get started with Sage, users can install it as a plugin directly from the GitHub repository for Claude Code, or use a VS Code extension package for Cursor users. An npm package, @gendigital/sage-openclaw, is also available for OpenClaw installations.

Conclusion and Call to Action

In conclusion, Sage is an essential tool for securing AI agents and preventing potential security threats. Therefore, we recommend installing Sage and taking advantage of its features to protect your developer workstations. Additionally, stay informed about the latest open-source cybersecurity tools by subscribing to our monthly newsletter.

Frequently Asked Questions

1. What is Sage and what does it do? Sage is an open-source tool that inserts an interception layer between AI agents and their operations, checking each action before it proceeds.
2. How does Sage work? Sage works through hook systems native to the agent platforms it supports, intercepting tool calls and passing them through several detection layers.
3. What are the benefits of using Sage? The benefits of using Sage include preventing potential security threats, protecting developer workstations, and staying informed about the latest open-source cybersecurity tools.
4. How can I install Sage? Sage can be installed as a plugin directly from the GitHub repository for Claude Code, or using a VS Code extension package for Cursor users. An npm package, @gendigital/sage-openclaw, is also available for OpenClaw installations.
5. What is the focus keyword for this article? The focus keyword for this article is AI agent security.