agent-native-reviewer

Agent-native parity ensures any application action accessible to users is also available to AI agents. It enforces consistent capabilities by verifying that UI actions, data context, and workspace sharing align between human workflows and agent tools.

You verify agent tool access by reviewing UI workflows and confirming each action has a corresponding agent tool. This action parity check identifies discrepancies where users can perform tasks that agents cannot execute.

Reviewing agent tools involves analyzing tool design to ensure they function as primitives rather than encoded business logic. This validates that tools provide foundational operations agents can compose, rather than hardcoding specific user workflows.

Context starvation occurs when agents lack access to the same data context as users. Analyzing system prompts and workspace sharing configurations identifies this anti-pattern and rectifies discrepancies in data context parity.

You check for sandbox isolation by analyzing application designs to identify when agents are restricted from accessing necessary workspace data. This review rectifies context discrepancies and validates proper agent-native architecture.

Yes, you can use code review to enforce agent-native principles after adding new features. Reviewing new UI workflows like report creation confirms agents have corresponding tools, maintaining parity between user and agent capabilities.