Critical Vulnerability Exposes GitHub Agentic Workflows to Prompt Injection

Researchers have discovered a critical vulnerability that allows attackers to use a specially crafted public GitHub Issue to deceive AI-powered workflows into revealing data from private repositories. This can be done without the need for authentication, potentially exposing sensitive information. The vulnerability is related to prompt injection in GitHub Agentic Workflows. Attackers can exploit this weakness to gain unauthorized access to private repository data. The issue highlights a potential security risk for users relying on AI-powered workflows in GitHub.