This article introduces shiftGrid, an open-source project designed to automate security testing processes using AI agents.The tool acts as a prompt engine for pentesting, allowing users to direct their agents through its API while maintaining human oversight.Key features include scope definition, checklist adherence, and real-time tracking of test observations and findings.

The platform provides transparency and traceability by generating detailed notes and findings lists, enabling operators to monitor agent activities closely.shiftGrid's extended context window not only identifies security vulnerabilities but also offers a comprehensive overview of applications and servers.Originally developed for security testing, the tool demonstrates versatility by potentially serving as a harness for various tasks.The author highlights its effectiveness in auditing systems like Confluence, emphasizing its ability to reveal critical insights.

By integrating AI agents with human guidance, shiftGrid represents an innovative approach to enhancing security assessments while maintaining operational control.