Worked on enhancing test coverage for prompt guard functionality within the meta-llama/PurpleLlama repository by introducing a new field, caught_by_promptguard, to the test case data model in query_llm.py. This addition allowed for more granular tracking of prompt guard scenarios, improving data fidelity and reducing the risk of undetected issues in production prompts. The work involved Python programming, data structure design, and careful test-data modeling to ensure backward compatibility with existing test structures. Leveraged Git for change tracking and focused on maintainability, enabling faster test iteration and more robust validation of prompt guard behavior without introducing integration risks.