Developed an agent security policy enhancement for the All-Hands-AI/agent-sdk repository, focusing on preventing the relocation or copying of files containing sensitive credentials or personal data, even during bulk operations. The solution was implemented in Python and involved updating both tests and snapshots to ensure the new security constraints were thoroughly validated. By introducing a secret-disclosure consent rule, the work addressed enterprise data protection requirements and aligned with data governance standards. The approach emphasized robust policy enforcement and comprehensive testing, resulting in a more secure handling of sensitive information within the codebase and reducing the risk of accidental data leakage.