Bitdefender finds GitHub malware that plants instructions in AI coding agents' project files to run itself and spread
Bitdefender Labs · Company report
Read the source (opens in a new tab)- What happened
- Bitdefender Labs said on November 17, 2026, that one of its researchers had identified 22 malicious GitHub links with files designed to make AI-assisted development tools run a concealed payload.
- Key points
- The campaign abuses agent instruction files such as AGENTS.md and CLAUDE.md, session-start hooks, MCP configurations and editor tasks; a hook can run the payload with no AI prompt at all.
- The payload can steal developer credentials, cloud secrets, CI/CD tokens, package-registry credentials, browser data and password-manager data.
- It searches the victim's machine for AI-agent configurations to infect, and can use stolen GitHub and GitLab tokens to plant malicious configurations in other repositories.
- Bitdefender says instruction files "should be reviewed with the same care" developers give dependencies, CI/CD workflows and scripts before opening an unfamiliar repository.
- Evidence
- A security vendor's analysis of malware it found, ending in a pitch for its own product. It does not say how many developers were infected or who is behind the campaign.