Audits application code and architecture to ensure AI agents have full feature parity with human users.
The Agent-Native Reviewer is a specialized diagnostic tool designed for developers building compound engineering systems where AI agents and humans collaborate. It performs deep architectural audits to ensure that every UI action has a corresponding agent tool, that agents share the same data context as users, and that system prompts are dynamically injected with runtime state. By identifying anti-patterns like context starvation, sandbox isolation, or 'orphan features,' this skill helps teams build applications where AI agents function as first-class citizens with full operational capabilities.
主な機能
01Validation of tool design based on modular primitives rather than hardcoded workflows
02Detection of architectural anti-patterns like Silent Actions and Sandbox Isolation
03Deep context parity audits for system prompt completeness and runtime state
04Automated action parity mapping between UI components and agent tools
0535 GitHub stars
06Mobile-specific checks for background execution and permission handling
ユースケース
01Refactoring complex, brittle agent tools into flexible primitives for better AI reasoning
02Auditing legacy codebases to transform traditional applications into agent-native platforms
03Reviewing Pull Requests that add new UI features to ensure immediate agent accessibility
