iFixAi
Command-line auditor that checks whether an AI agent does its intended job
What it is
A Python diagnostic that runs adversarial and assurance checks against an AI agent or model endpoint, then scores whether it behaves according to its role, permissions, and business rules. It can be driven by a setup wizard, explicit CLI flags for CI, or as a plugin or skill inside Claude Code, Codex, and other agents, with grading by the model itself, an independent vendor, or a panel of judges. Teams shipping agents into production use it to surface issues such as unauthorized actions, hallucinations, and prompt injection, and get JSON and Markdown reports.