The New Stack reported that a malicious pull request nearly converted an AI coding assistant into a wiper. The article stresses that AI agents need strict external security controls to avoid exploitation. It credits env zero for the analysis and presents the event as a warning about unvetted code changes in AI-assisted development. The piece frames the risk as one where a single request could enable destructive actions if agents operate without isolation. According to the source the incident shows why safeguards must be placed outside the agent itself rather than relying on internal checks alone.