Aikido Security has identified a vulnerability in GitHub Actions and GitLab CI/CD workflows that allows AI agents to execute malicious instructions, potentially leaking sensitive information. The flaw affects multiple companies and demonstrates how AI prompt injection can compromise software supply chains.

This article details LinkedIn's efforts to upgrade its Static Application Security Testing (SAST) capabilities. It covers the challenges faced with legacy systems, the design principles guiding the modernization, and the implementation of a new GitHub Actions-based workflow to enhance security without disrupting developer productivity.

Recent security issues with GitHub Actions involved the modification of the tj-actions/changed-files action, which leaked secrets. To mitigate such vulnerabilities, the author recommends pinning versions to explicit commit hashes and introduces a tool called "just an installer" (jas) that verifies downloads using SHA-256 hashes, enhancing the reliability of binary installations in GitHub Actions workflows.

Researchers discovered vulnerabilities in the Nix ecosystem related to GitHub Actions, specifically concerning the pull_request_target event, which could allow for supply chain attacks and command injection. They identified two significant flaws: one involving xargs and the other enabling symbolic link exploitation, leading to unauthorized access to sensitive data. The maintainers acted quickly to disable the vulnerable workflows and implement fixes.

Grafana Labs introduced Zizmor, an open source static analysis tool, in their CI/CD pipelines to detect and prevent vulnerabilities in GitHub Actions following a security incident. The tool helps identify unsafe configurations and practices, such as the use of `pull_request_target`, and is part of a broader effort to enhance security across their repositories. Despite facing challenges like GitHub's rate limiting, Grafana is committed to using Zizmor to bolster their defenses against future attacks.

The article discusses the security considerations necessary for using GitHub Actions in CI/CD setups, emphasizing the importance of protecting workflows against potential threats from contributors with write access. It details various attack scenarios, including script injection vulnerabilities, and provides best practices for securing sensitive workflows and managing permissions effectively.

GitHub Actions allows users to specify any executable on the $PATH as the shell for running commands, rather than being limited to predefined shell values. This flexibility enables various unconventional uses, such as executing C code directly or modifying the $PATH dynamically. However, this raises potential security concerns, as it allows for unexpected behaviors in the execution environment.

Sysdig's Threat Research Team uncovered significant security vulnerabilities in GitHub Actions workflows across popular open source projects, including those by MITRE and Splunk. Their research revealed how insecure configurations, particularly using pull_request_target, can expose sensitive credentials and allow for exploitation, prompting the team to recommend best practices to enhance CI/CD security.