Show HN: Amidumb – scores how you direct Claude Code and Codex, from local logs (github.com)

Amidumb has launched a unique tool that evaluates how effectively developers interact with AI coding agents like Claude Code and Codex. By analyzing the command history stored locally, Amidumb scores users on a scale from 0 to 100, assessing five key aspects: specification, steering, verification, discipline, and leverage. This feedback includes a humorous title based on the score, ranging from "Agent's Worst Nightmare" to "Hostage," as well as a roast from the AI itself based on users' past prompts. Notably, the tool runs entirely locally, requires no sign-up or API keys, and guarantees that user data remains private. The significance of Amidumb lies in its potential to enhance AI collaboration skills among developers, offering a tangible way to improve their interaction with AI agents. By providing detailed sub-scores and actionable insights, users can identify areas for improvement in their coding approach. Its local operation and redaction measures ensure security while allowing for rich feedback, making it a valuable addition to the toolkit of engineers working in AI/ML environments. This innovative approach fosters a more disciplined and effective use of AI coding assistants, ultimately driving productivity and improving outcomes for software development.