Worked on enhancing the security of the Hermes Agent approval system in the NousResearch/hermes-agent repository by developing detection patterns for potential abuse scenarios. The solution targeted vulnerabilities such as heredoc script injection, PID expansion self-termination using kill commands, destructive Git operations, and the two-step chmod +x followed by execution pattern. Leveraging Python and Shell scripting, the work included comprehensive automated tests to validate each detection pattern and ensure robust coverage. This feature addressed four security gaps identified through a source-grounded audit, demonstrating a methodical approach to security-focused Python development and testing within a collaborative open-source environment.