How to Audit All Python Virtual Environments for Compromised Packages Without Executing Python

The growing number of supply-chain attacks targeting popular packages has prompted developers to seek reliable ways to audit their Python installations. Recent incidents affected the npm package Axios, the vulnerability scanner Trivy, and the Python package LiteLLM. In response, one developer created a system to locate every virtual environment on a machine and inspect its installed packages without ever executing Python code.

Virtual environments isolate project dependencies by storing symbolic links to the Python interpreter and site-packages inside a dedicated directory, conventionally named .venv. Because Python itself does not track created environments, the author maintains a personal registry file at ~/.venv_registry that is updated both by periodic find commands and by a shell function that records every newly created environment.

The registry enables a simple Bash script that reads each path, verifies the directory still exists, and then calls uv pip freeze --python against the environment’s interpreter. Because uv is written in Rust, it parses installation metadata without running any Python code—an important safeguard after the LiteLLM compromise demonstrated that even innocuous commands such as python --version could execute attacker-controlled .pth files.

The same registry supports additional security and maintenance tasks. It can be used to identify outdated package versions across all projects, locate remaining usages of libraries the developer wishes to retire, and search project source trees for calls to specific functions without scanning the entire disk.

Global installations outside virtual environments are prevented by setting PIP_REQUIRE_VIRTUALENV=true and by relying on uv, which refuses to modify the system Python unless the --system flag is explicitly supplied. The resulting workflow therefore provides both rapid detection of known malicious packages and ongoing hygiene for a large collection of Python projects on macOS.