An issue in the Planner Agent component allows unauthenticated remote attackers to execute arbitrary code. The flaw exists in the `ExecutionAction.run()` function within the `mindsearch/agent/graph.py` file, where the `inputs` argument is processed without adequate validation or sanitization. This allows an attacker to inject and execute arbitrary Python code, potentially granting full control over the application instance, enabling data exfiltration, system manipulation, and lateral movement within the hosting environment.