dk-agent-native-audit

Agent-native architecture ensures codebases support AI agents through principles like action parity and context injection. You audit it by running parallel sub-agents that evaluate your code against eight core principles, generating a scored report with identified gaps and recommendations.

You can perform an agent-native code review by invoking a parallelized audit that scores your codebase across eight principles, including CRUD completeness, capability discovery, and UI integration, producing a final report with actionable improvement recommendations.

Yes, you can invoke the architecture audit with an optional principle argument to limit the scope, allowing you to perform a targeted review of a single area like context injection instead of running all eight parallel assessments.

The eight principles evaluated are Action Parity, Tools as Primitives, Context Injection, Shared Workspace, CRUD Completeness, UI Integration, Capability Discovery, and Prompt-Native Features, each scored by dedicated parallel sub-agents.

Multi-agent parallel auditing works by launching dedicated sub-agents simultaneously, each assessing a specific architecture principle, which generates per-principle scores that are compiled into a comprehensive final report with overall scores and recommendations.

Yes, the audit produces a final report that includes overall scores, identified architectural gaps across the eight principles, and actionable recommendations to guide improvements in your codebase.