A coordinated campaign targeted automation workflows across 61 public GitHub repositories, including widely used development and security projects. Attackers placed hidden commands inside issues, pull requests or comments so vulnerable workflows would execute them automatically. Researchers identified 83 malicious items and found two cases where commands ran inside the automation environment. They could not confirm that credentials were stolen, source code was changed or downstream software was compromised.
Why It Matters
A public contribution channel can become an execution path when workflow code inserts untrusted content into shell commands. For leaders: open-source governance must include the automation surrounding code, not only code review itself. For defenders: isolate externally triggered jobs, limit available secrets and environment variables, and apply least privilege to workflow permissions. Maintainers should treat all issue and pull-request content as hostile input. The larger lesson is that convenience-driven automation can quietly expand the software supply-chain attack surface.
What Defenders Should Do
Review externally triggered workflows now and remove direct use of untrusted content from privileged shell execution.