GitHub threat report

Researchers at Noma Labs disclosed GitLost, a critical prompt injection vulnerability in GitHub's Agentic Workflows that allows unauthenticated attackers to exfiltrate sensitive data from private repositories. The flaw exploits insufficient trust boundary separation between system instructions and user-controlled input in GitHub Issues. Attackers can craft malicious issue content to trick the AI agent into executing unauthorized commands and leaking private repository data as public comments.

• GitLost Vulnerability Lets Attackers Trick GitHub AI Agent Into (opens in a new tab)
• Critical Vulnerability Exposes GitHub Agentic Workflows to Promp(opens in a new tab)