GitHub radar
iFixAi: An AI agent audit in less than 2 minutes!
iFixAi evaluates AI agents from A to F on 5 key pillars with 49 checks — as a Claude Code plugin/Codex plugin/CLI wizard/One Click Skill for Cursor/VS Code/Windsurf/Cline/… Result is generated within 2 mins. Launched in late April 2026, it has gained over 11,000 stars within 4 months.
iFixAi is a standalone AI agent evaluation system that focuses on non-technical aspects such as deception, fabrication, manipulation, unpredictability, opacity etc. It can be used by any AI service providers including OpenAI, Anthropic, Gemini, Azure, Bedrock, OpenRouter, or custom AI agents via HTTP interface. It provides a grade from A to F along with detailed breakdown on each pillar. You can evaluate any AI agent whether it’s from OpenAI, Anthropic, Gemini, Azure, Bedrock, OpenRouter, or even your own AI agent. It can also be used as a Claude Code plugin/Codex plugin/CLI wizard/One Click Skill for Cursor/VS Code/Windsurf/Cline/… When it comes to citable grade, we need a second opinion from another independent AI service provider (not the one we are evaluating). It takes less than 2 minutes to run all the tests. Approximately $12–18 if we use claude-sonnet-4.6 as the judge (OpenRouter price in mid-2026). Licensed under Apache 2.0.
Why a vibe-coder should care
Current AI evaluation systems only focus on technical aspects such as efficiency, latency, prompt injection etc. But none of them tell us if the AI agent is performing well in its intended task. iFixAi is designed to evaluate AI agents at a higher level, provide a quantitative score (from A to F) and explain why it received a low score. It can be done within 2 minutes via plugins or command lines.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Install the iFixAi plugin from https://github.com/ifixai-ai/iFixAi in Claude Code (commands: `/plugin marketplace add ifixai-ai/iFixAi` then `/plugin install ifixai@ifixai-community`), run an audit of my agent, and show me the scorecard with per-pillar breakdown. Ask me for the API key for the provider you need.
Runs on a regular laptop — only Python 3.10+ and an API key from your chosen provider needed.
Open on GitHub▌ More finds