agent-auditor

To audit AI agents in a Railway PostgreSQL database, connect to the database and pull all agents to inspect their system prompts and behavior. The audit scores every agent on a 10-dimension rubric and returns a summary of enterprise quality, needs improvement, and critical gaps.

The 10-dimension rubric is an enterprise quality standard used to score AI agents on adherence to required roles, output formats, constraints, and failure modes. It identifies missing dimensions and flags any agents scoring below 8 out of 10 for improvement.

To generate an improvement plan for agents scoring below 8/10, run the audit to evaluate the 10-dimension rubric. The process produces a prioritized plan with concrete additions for missing roles, output formats, constraints, and failure modes, plus guidance to apply changes in Railway.

Yes, you can use this audit as a pre-deployment quality gate. It produces a comprehensive enterprise-quality report that identifies critical gaps and missing dimensions before deployment, ensuring adherence to the 2026 standard and driving continuous improvement across the agent fleet.

You need access to your Railway PostgreSQL database to run the agent quality audit. Once connected, the process automatically pulls all agents to inspect their system prompts and behavior, scoring them against the 10-dimension rubric without requiring additional dependencies.

The audit is limited to inspecting agents stored within the Railway PostgreSQL database. It evaluates system prompts and behavior against the 10-dimension rubric, but applying the generated improvement plan requires a manual review before changes are pushed to Railway.