Credential-Stealing GitHub Actions Attack Hits 340 Repositories

Cybersecurity researchers have uncovered a large‑scale credential‑theft campaign that leveraged two high‑profile open‑source maintainer accounts to push a malicious workflow into hundreds of repositories on GitHub. The operation began on October 9, 2026 when an attacker logged in as Takashi Kitao—author of the 18,400‑star game engine pyxel—and deployed a harmful workflow to 27 repositories starting at 13:20 UTC ¹.

Within hours, the same attacker expanded the reach to over 340 repositories across various organizations. The malicious script could read and exfiltrate secrets stored within the repository environment, possibly leaking API keys, passwords, or tokens used by developers in CI/CD pipelines. This scale demonstrates how attackers can exploit trust relationships among third‑party integrations such as those found at GitHub’s marketplace ¹.

The vulnerability lies in the fact that many teams use CI/CD workflows built on platforms like GitHub. These workflows often contain sensitive credentials; if an attacker gains control over a workflow they can inject arbitrary code into builds, exfiltrating data from the repository environment. Similar attacks were seen before—in 2019 a compromised npm package named “coa” stole credentials from more than four thousand applications.

This new campaign shows an evolution toward more targeted supply‑chain compromise and highlights weaknesses across software development practices that require better authentication controls for third‑party integrations such as those found at Github’s marketplace.

Experts note that the core issue is not just the presence of secrets but also the lack of robust multi‑factor authentication for account access. Attackers may be using leaked credentials from previous breaches as part of their chain, making it harder for defenders to detect unauthorized activity until after deployment. In practice, organizations should enforce least‑privilege principles, rotate secrets regularly, and employ secret scanning tools that flag hard‑coded values in repositories. They should also consider moving away from storing secrets directly in workflow files and instead use encrypted vaults or dedicated secret management services with fine‑grained access policies.

While the immediate impact has been credential theft, the broader implication is a warning about the trust inherent within CI/CD pipelines. If attackers can gain footholds through legitimate maintainer accounts, any organization relying on open‑source components could become vulnerable without noticing. The incident underscores the need for tighter security around third‑party integrations and for developers to audit the provenance of packages they incorporate into their builds.

The attack demonstrates how quickly malicious code can spread when trusted maintainers are compromised—an alarming reminder that even well‑established projects like pyxel can serve as vectors for large‑scale attacks if proper safeguards are absent. As researchers continue to investigate the full extent of the breach, industry stakeholders must review current practices and adopt stronger controls over who can modify workflows and where secrets reside.