The Gheist campaign is actively exploiting compromised GitHub tokens to inject malicious workflows into repositories, aiming to exfiltrate sensitive secrets and credentials from private and enterprise environments.
Key Points
- Since September 15, 2026, attackers have used the "gheist" user agent to plant automated "sync-check" workflows in over 850 compromised GitHub accounts.
- The malicious workflows bypass standard security by printing secrets to logs, which are then exfiltrated via the GitHub API to avoid external network detection.
- While GitHub’s approval requirements block these attacks in public repositories, private repositories and enterprise environments remain vulnerable to unauthorized workflow execution.
- Advanced variants of the tool attempt to compromise self-hosted runners by searching for private keys, cloud credentials, and tokens for AI providers like Anthropic and NVIDIA.
- The campaign overlaps with other malicious activities, including cryptomining, malware distribution, and the use of infostealers to harvest initial access credentials.