The TrustFall research by Adversa AI demonstrates that agentic AI coding tools can be tricked into executing malicious code from cloned repositories with minimal user interaction, exploiting the inherited convention of trusting repository contents. The article does not provide specific CVSS scores, affected version ranges, fixed versions, or workarounds for the covered tools.
Developers clone unfamiliar repositories all the time. Open-source projects, work from teammates, sample code from a tutorial, a library someone recommended on a forum. The convention is old and reasonable: you look at what’s inside before you run it. AI coding assistants that work from the command line have inherited that convention, and a new piece of research from Adversa AI shows where the convention breaks. The research, called TrustFall, covers four agentic coding tools: … More → The post One keypress is all it takes to compromise four AI coding tools appeared first on Help Net Security .