agent-native-reviewer

Agent-native parity ensures any action a user can perform within an application is also accessible to an AI agent. It focuses on maintaining equivalent capabilities between human users and agents regarding action parity, context parity, and shared workspace usage.

You verify action parity by reviewing code and application designs to check if UI actions have corresponding agent tools. This ensures your AI assistant can perform new features through its available tools, not just through the user interface.

You check context parity by analyzing application designs to verify that agents have access to the same data as users. This involves reviewing dynamic context injection mechanisms to ensure AI-driven applications receive equivalent shared workspace data.

Yes, you can use this review process for existing web applications. After adding a new feature, run the review to confirm your AI assistant can perform the same action through its available primitive tools, maintaining strict action parity.

The review focuses strictly on agent-native principles like action parity, context parity, shared workspace, primitive tools, and dynamic context injection. It does not cover general code quality, security vulnerabilities, or performance optimization outside agent capability equivalence.