Developed tool-calling support for the LassoGuardrail component in the BerriAI/litellm repository, enabling secure and robust tool interactions with comprehensive input and output masking. Focused on backend and API development using Python, the work included refactoring duplicated access patterns into shared helpers to streamline code and improve maintainability. Enhanced reliability by expanding end-to-end and edge-case test coverage, ensuring the pipeline handled multimodal tool calls and sensitive data correctly. Improvements to continuous integration incorporated black formatting and mypy typing, raising code quality standards. The solution addressed data leakage risks and strengthened automated guardrails for tool-calling workflows within the repository.