Worked on security hardening for the meta-llama/PurpleLlama repository by addressing a Remote Code Execution vulnerability in the Canary Exploit Verification component. The solution involved replacing the use of eval with ast.literal_eval to safely evaluate byte strings, thereby reducing the risk of malicious code execution at runtime. This change was implemented in Python and focused on improving runtime safety without altering the public API or affecting external behavior. The work demonstrated skills in Python programming, cybersecurity, and vulnerability assessment, and was completed through a peer-reviewed process before being merged, ensuring robust mitigation of the identified security issue.