Agentic AI Security Testing

In this module of the Advanced AI Security Part 1 learning path, you will practice agentic AI security testing against the agent loop. Persona is not IAM. A goal string that contains a relative path is enough for the lab agent to read sibling secrets. Upon completion of the module, students will be able to: - Spoof a persona and show it does not change authorization. - Drive read_workspace from /agent/run without calling /mcp yourself. - Cluster the finding with the MCP traversal as one root cause, two entrypoints. For inquiries, please write us at https://www.darkrelay.com/cybersecurity-course-inquiry