How a Downloads Folder File Can Hijack Your AI Coding Agent

Forever Security discovered an AI coding agent installing a malicious npm package due to a plain text file in a customer's Downloads folder. This attack required no exploit or CVE, as agents trust files in directories where websites can write. Separately, implementing auditable agent memory is challenging because vector stores often lack records of used memories. Using ZizkaDB's search and context functions allows developers to build inspectable memory, ensuring every agent answer is traceable to specific records.