Ensures the customer support agent never crosses into regulated financial advisory territory. Prohibits personalized investment guidance, suitability assessments, and market predictions regardless of how a user frames their request.

Protect system integrity and organizational security by establishing clear technical and ethical boundaries for automated coding interactions. This policy prevents unauthorized access and ensures responsible computational resource management.

Governs the behavior of the sales lead agent to protect confidential business data, ensure legal compliance with communication regulations, and maintain professional standards across all lead interactions.

Create a governance framework that protects employee privacy and ensures organizational fairness through rigorous ethical standards. This policy establishes boundaries around confidentiality and transparent decision-making to maintain both individual rights and organizational integrity.

Ensures HIPAA compliance and safe patient interactions for front-desk and scheduling agents. Protects sensitive health information while maintaining efficient care coordination.

Enforces HR compliance, prevents disclosure of sensitive employee data, and ensures fair and unbiased handling of all personnel-related interactions.

Governs research agents to ensure accuracy, source integrity, and responsible handling of sensitive or contested information during automated research tasks.

Controls calendar access, scheduling permissions, and attendee management for automated scheduling agents. Ensures user consent and prevents unauthorized calendar modifications.

Defines privacy boundaries, action scope, and authorization requirements for personal assistant agents. Ensures the agent acts only on explicit instructions and protects user data across sessions.

A foundational content safety baseline covering the most critical harm categories. Prevents output related to weapons of mass destruction, child exploitation, defamation, fraud, illegal surveillance, violent crime, IP violations, privacy breaches, and self-harm.

A minimal baseline policy requiring the assistant to strictly adhere to the rules provided in the system prompt.

A safety baseline for agents with shell access. Prevents dangerous or harmful bash commands that could render the system inoperable, vulnerable, or leak sensitive information.

A minimal safety baseline for coding agents with shell access. Prevents dangerous or harmful bash commands that could render the system inoperable, vulnerable, or leak sensitive information.

These scenarios comprise of various scenarios that may arise with Agent Alpha usecases. This group includes scenarios around illegal financial advice, money laundering, etc.