safedep/pmg

PMG protects developers, AI agents from malicious open source packages using proxy, sandbox and SafeDep's threat intelligence feed.

What it solves

Package Manager Guard (PMG) addresses the risk of malicious code execution during software development. Developers and AI coding agents frequently install thousands of lines of code via package managers like npm or pip without manual review. PMG prevents supply chain attacks by intercepting these installs and blocking malicious packages before they execute on your system.

How it works

PMG acts as a transparent security layer that wraps your existing package manager commands. It employs a defense-in-depth strategy:

  1. Threat Intelligence: It checks every package against a real-time database of known malicious threats.
  2. Dependency Cooldown: It blocks recently published package versions to mitigate risks from newly compromised software.
  3. Sandboxing: It can optionally run installation scripts within OS-native sandboxes (such as macOS Seatbelt or Linux Landlock) to restrict system access.
  4. Audit Logging: It maintains a local log of all package installation activity for security visibility.

Who it’s for

It is designed for individual developers, security-conscious teams, and AI coding agents that need to safely automate dependency management without changing their existing workflows.

Highlights

  • Zero-config integration: Works seamlessly with common package managers like npm, pip, poetry, uv, and cargo.
  • Transparent protection: No changes required to your existing terminal commands or AI agent workflows.
  • No account required: Free, open-source, and does not require API keys or registration.
  • Advanced security: Includes optional kernel-level enforcement and OS-native sandboxing for high-security environments.

Related

  • Dispatch
    Leveraging Frontier Models for Package Registry Malware Detection

    An exploration of the feasibility and challenges of using LLMs to secure the software supply chain against malicious packages in registries like NPM.

  • Project
    GoPlusSecurity/agentguard

    A real-time security layer for AI agents that blocks destructive commands, scans plugins for malicious code, and performs daily security audits to prevent data theft and system damage.

  • Project
    canyonroad/agentsh

    A secure, policy-enforced execution gateway for AI agents that intercepts and governs file, network, process, and database activity at runtime.

  • Project
    akitaonrails/ai-jail

    A CLI that runs AI coding agents in an OS sandbox (bubblewrap/Landlock/seccomp on Linux, sandbox-exec on macOS), scoped to prevent accidents — not to stop attackers. The project directory is read-write; the host is hidden by default, with opt-in flags for credentials, network, GPU, display, etc. Includes a strict `--lockdown` mode and a "phantom credential" proxy that keeps real API keys out of the sandbox. A mature, well-documented Rust project — not a list, tutorial, or demo.

  • Project
    PrismorSec/prismor

    Prismor is a runtime security and governance platform for AI agents that monitors and enforces policies on agent actions, tool calls, and network access to prevent security risks like prompt injection and data exfiltration.